Coronavirus disease 2019 (COVID-19) has spread to >500 million people and caused 6.2 million deaths worldwide as of 22 April 2022.1 Approximately 20% of patients with COVID-19 develop severe symptoms and 5% of patients require intensive care.2 A wide range of complications were reported with COVID-19 infection, including nervous system diseases,3 4 circulatory system diseases,5 6 7 8 9 urinary system diseases,10 and digestive system diseases.11

Various genetic associations with COVID-19 outcomes have been explored.12 13 14 15 A whole-genome sequencing study of germline mutations revealed a cluster of six genes (SLC6A20, CCR9, FYCO1, CXCR6, XCR1, and LZTFL1) that increased susceptibility to severe COVID-19 with respiratory failure.16 In a Chinese population, a whole-genome sequencing study of 332 patients with COVID-19 identified loci in the genes TMEM189 and UBE2V1 with potential genome-wide implications through the IL-1 signalling pathway.17 In an intensive care unit cohort of 15 patients with severe COVID-19, analysis of RNA sequencing (RNA-Seq) data from blood samples showed that the immune-modulating genes PD-L1 and PD-L2 were differentially expressed among patients with fatal outcomes.18

Thus far, studies of gene expression at initial sites of infection in patients with severe and non-severe COVID-19 remain limited. To investigate COVID-19 from a genetic perspective, we conducted RNA-Seq analysis of respiratory tract samples from patients with COVID-19; we sought to identify genes associated with disease severity.

Twenty-four patients were recruited from Prince of Wales Hospital in Hong Kong between 7 February and 10 April 2020. All patients had severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, confirmed by two independent real-time reverse transcription–polymerase chain reaction assays targeting the N gene.19 Symptoms on admission were recorded, and medical histories were collected from clinical health records. Among the recruited patients, eight who received mechanical ventilation or extracorporeal membrane oxygenation were regarded as severe cases; the remaining 16 patients showed asymptomatic or mild (no pneumonia) to moderate (pneumonia but not requiring oxygen supplementation) disease and were regarded as non-severe cases. RNA sequencing was performed on upper and lower respiratory swab samples collected within 3 days after hospitalisation.

Total RNA was extracted from respiratory swab samples using the QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany), pre-treated with DNase I and depleted of human rRNA and globin genes using the QIAseq FastSelect ribosomal RNA and globin mRNA Removal Kit (Qiagen, Hilden, Germany). Illumina libraries for RNA-Seq next-generation sequencing were prepared using the KAPA HyperPrep Kit (Roche, Pleasanton [CA], US) in accordance with the manufacturer’s instructions, then sequenced on an Illumina NextSeq 500 system (Illumina, San Diego [CA], US) using 150 bp paired-end reads. The raw data consisted of 58 735 Ensembl-annotated20 genes. Quality control was performed to remove patients with low numbers of RNA-Seq reads (three samples) and genes with zero reads in >20% of samples (55 571 genes). Thus, 3164 genes remained available for differential expression analysis. Raw read count data were summarised as fragments per million reads mapped21 and then log₂-transformed.

Logistic regression was utilised to identify genes associated with severity outcomes. Subsequent evaluations by the Kolmogorov–Smirnov (KS) test were performed to test for differences in gene expression between groups. The Bonferroni corrected significance threshold was 1.58 × 10^-5.

According to their disease relevance in the GeneCards21 22 and MalaCards23 24 databases, genes were categorised into the following 10 disease groups: nervous system, integumentary system, circulatory system, urinary system, digestive system, respiratory system, musculoskeletal system, endocrine system, reproductive system, and infectious diseases.

The patients’ demographics and baseline characteristics are summarised in Table 1. The mean age of patients in the severe group was 62.13 years (95% confidence interval [CI]=53.34-70.91), which was significantly higher than that in the non-severe group (29.73 years; 95% CI=22.11-37.36). Compared with the non-severe group, the severe group had higher prevalences of complications including cardiovascular, liver, endocrine, and metabolic disorders, as well as higher rates of respiratory, fever, and diarrhoea symptoms. The COVID-19 World Health Organization score25 was significantly higher in the severe group than in the non-severe group. Lopinavir, antibiotics, conventional oxygen therapy, and mechanical ventilation were more commonly used for treatment in the severe group than in the non-severe group (Table 1).

Six genes, namely RPL15, BACE1-AS, CEPT1, EIF4G1, TMEM91, and TBCK, were differentially expressed between the severe and non-severe groups (all P values <0.05 in both logistic regression and the KS test) [Table 2]. Fold-change and odds ratio results indicated that these genes were consistently highly expressed in the severe group. The complete list of genes with P values <0.05 in KS test is provided in online supplementary Table 1.

The functions of the identified genes were summarised through database and literature searches. Two genes, 60S ribosomal protein L15 (RPL15) and eukaryotic translation initiation factor 4 gamma 1 (EIF4G1), play roles in host translation of viral mRNA.26 27 28 Furthermore, the top genes were mainly involved in neurological disorders. RPL15 is involved in the life cycle of human immunodeficiency virus,29 30 and baculovirus infection reportedly disrupts the expression of this gene.31 32 EIF4G1 plays a role in viral binding and affects the pathogenicity and virulence of H5N1 influenza A virus, foot-and-mouth disease virus, and vaccinia virus33 34 35; it also contains multiple mutations among patients with familial Parkinson’s disease.36 TBCK encodes a conserved protein kinase that regulates cell size and proliferation.37 CEPT1 encodes choline/ethanolamine phosphotransferase, which is used in the synthesis of choline- or ethanolamine-containing phospholipids. The function of TMEM91, a transmembrane protein, is unclear; however, the results of genome-wide association studies suggest that loci containing this gene are involved in lung diseases.

The non-coding gene BACE1-AS regulates the stability of the BACE1 protein and directly increases the abundance of amyloid beta-peptide (Aβ_1-42) in Alzheimer’s disease.38 The implications of this gene in severe COVID-19 are unclear. For the top 15 overexpressed genes (P values in KS test <0.05), disease relevance data were retrieved from GeneCards22; 14 of the 15 genes (93.3%) have been linked to neurological diseases, followed by eye (80.0%) and psychiatric (73.3%) diseases. Thus, all of the top genes were involved in nervous system disorders (Fig).

Nervous system disorders such as encephalopathy, impaired consciousness, seizure, ataxia, neuropathies, neurodegenerative diseases, and anosmia have been extensively documented in patients with COVID-19.39 40 41 The two major potential pathogenesis pathways are direct viral invasion and immune-mediated injury. Direct viral entry to the central nervous system can travel through hematogenous or olfactory routes, or by transneuronal spread from the lungs.42 Post-mortem analysis of brain tissue from patients with COVID-19 encephalitis reportedly contained SARS-CoV-2 viral particles.43 44 Furthermore, a series of autopsy studies showed that localised inflammation of the brainstem nuclei, as well as the cytokine storm associated with SARS-CoV-2 infection, could disrupt the blood–brain barrier and cause necrosis in the brains of patients with severe COVID-19.45 46 In patients with COVID-19, anosmia may be caused by an inflammation-mediated decrease in odorant receptor expression.47 Several studies have utilised RNA-Seq to characterise the transcriptomic profiles of patients with COVID-19.48 49 Significant downregulation of genes related to the hypoxia-inducible factor system was observed during periods of infection and oxygen deprivation.50 Additionally, transcriptomic profiles of peripheral blood mononuclear cells revealed that patients with COVID-19 shared several dysregulated genes with individuals who had bipolar illness, post-traumatic stress disorder, or schizophrenia.51 The present findings suggest that SARS-CoV-2 infection is associated with differential expression of genes involved in nervous system disorders. Future studies that involve gene expression profiling with larger sample sizes, in vitro infection experiments, and animal models can help to elucidate the mechanisms and corresponding therapeutic approaches for neurological complications of COVID-19.

A major limitation of this study was its small sample size. Patient age distributions considerably differed between groups. However, age-stratified analysis showed effects consistent with the directions reported in Table 2, although the statistical significance was hindered by the small sample size (online supplementary Table 2 and online supplementary Fig). Further sequencing of samples collected from respiratory tract sites may provide stronger evidence of protein expression abnormalities at the initial site of SARS-CoV-2 infection.

In this study, we conducted RNA-Seq analysis to identify differentially expressed genes between patients with severe and non-severe cases of COVID-19. We observed overexpression of genes involved in cell proliferation, viral binding and replication, and neurological and lung diseases, suggesting a pathophysiological mechanism by which SARS-CoV-2 induces lung inflammation and neurological complications.

Concept or design: BCY Zee, PKS Chan, MH Wang.
Acquisition of data: Z Chen, PKS Chan.
Analysis or interpretation of data: Q Li, Z Chen, Y Zhang, RWY Chan, MKC Chong, PKS Chan, G Lui, L Ling.
Drafting of the manuscript: Q Li, MH Wang.
Critical revision of the manuscript for important intellectual content: Q Li, Y Zhang, MH Wang.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

BCY Zee is a shareholder of Health View Bioanalytic Limited. As a statistical adviser of the journal, MKC Chong was not involved in the peer review process. MH Wang is a shareholder of Beth Bioinformatics Co, Ltd. Other authors have disclosed no conflicts of interest.

This research was partially supported by the Health and Medical Research Fund of the former Food and Health Bureau, Hong Kong SAR Government (Ref Nos.: COVID190103, COVID190112 and INF-CUHK-1), The Chinese University of Hong Kong (CUHK) Project Impact Enhancement Fund (Ref No.: CUPIEF/Ph2/COVID/06) and CUHK Direct Grant (Ref No.: 2020.025). The funders had no role in study design, data collection/analysis/interpretation, or manuscript preparation.

The study protocol of this research was approved by the Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee (Ref No.: 2020.076). All patients provided written informed consent for participation in this research.

The supplementary material was provided by the authors and some information may not have been peer reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by the Hong Kong Academy of Medicine and the Hong Kong Medical Association. The Hong Kong Academy of Medicine and the Hong Kong Medical Association disclaim all liability and responsibility arising from any reliance placed on the content.

1. World Health Organization. COVID-19 global data. Available from: https://covid19.who.int/WHO-COVID-19-global-data.csv. Accessed 22 Apr 2022.

2. Wu Z, McGoogan JM. Characteristics of and important lessons from the coronavirus disease 2019 (COVID-19) outbreak in China: summary of a report of 72 314 cases from the Chinese Center for Disease Control and Prevention. JAMA 2020;323:1239-42. Crossref

3. Gusev EI, Martynov MY, Boyko AN, et al. The novel coronavirus infection (COVID-19) and nervous system involvement: mechanisms of neurological disorders, clinical manifestations, and the organization of neurological care. Neurosci Behav Physiol 2021;51:147-54. Crossref

4. Kubota T, Kuroda N. Exacerbation of neurological symptoms and COVID-19 severity in patients with preexisting neurological disorders and COVID-19: a systematic review. Clin Neurol Neurosurg 2021;200:106349. Crossref

5. Zhou F, Yu T, Du R, et al. Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study. Lancet 2020;395:1054-62. Crossref

6. Li X, Xu S, Yu M, et al. Risk factors for severity and mortality in adult COVID-19 inpatients in Wuhan. J Allergy Clin Immunol 2020;146:110-8. Crossref

7. Chen R, Liang W, Jiang M, et al. Risk factors of fatal outcome in hospitalized subjects with coronavirus disease 2019 from a nationwide analysis in China. Chest 2020;158:97-105. Crossref

8. Docherty AB, Harrison EM, Green CA, et al. Features of 20 133 UK patients in hospital with COVID-19 using the ISARIC WHO Clinical Characterisation Protocol: prospective observational cohort study. BMJ 2020;369:m1985. Crossref

9. Richardson S, Hirsch JS, Narasimhan M, et al. Presenting characteristics, comorbidities, and outcomes among 5700 patients hospitalized with COVID-19 in the New York City area. JAMA 2020;323:2052-9. Crossref

10. Cheng Y, Luo R, Wang K, et al. Kidney disease is associated with in-hospital death of patients with COVID-19. Kidney Int 2020;97:829-38. Crossref

11. Mao R, Qiu Y, He JS, et al. Manifestations and prognosis of gastrointestinal and liver involvement in patients with COVID-19: a systematic review and meta-analysis. Lancet Gastroenterol Hepatol 2020;5:667-78. Crossref

12. Pairo-Castineira E, Clohisey S, Klaric L, et al. Genetic mechanisms of critical illness in COVID-19. Nature 2021;591:92-8. Crossref

13. COVID-19 Host Genetics Initiative. The COVID-19 Host Genetics Initiative, a global initiative to elucidate the role of host genetic factors in susceptibility and severity of the SARS-CoV-2 virus pandemic. Eur J Hum Genet 2020;28:715-8. Crossref

14. Anastassopoulou C, Gkizarioti Z, Patrinos GP, Tsakris A. Human genetic factors associated with susceptibility to SARS-CoV-2 infection and COVID-19 disease severity. Human Genomics 2020;14:40. Crossref

15. Kousathanas A, Pairo-Castineira E, Rawlik K, et al. Whole-genome sequencing reveals host factors underlying critical COVID-19. Nature 2022;607:97-103. Crossref

16. Severe Covid-19 GWAS Group; Ellinghaus D, Degenhardt F, et al. Genomewide association study of severe COVID-19 with respiratory failure. N Engl J Med 2020;383:1522-34. Crossref

17. Wang F, Huang S, Gao R, et al. Initial whole-genome sequencing and analysis of the host genetic contribution to COVID-19 severity and susceptibility. Cell Discov 2020;6:83. Crossref

18. Fredericks AM, Jentzsch MS, Cioffi WG, et al. Deep RNA sequencing of intensive care unit patients with COVID-19. Sci Rep 2022;12:15755. Crossref

19. Lui G, Ling L, Lai CK, et al. Viral dynamics of SARS-CoV-2 across a spectrum of disease severity in COVID-19. J Infect 2020;81:318-56. Crossref

20. Howe KL, Achuthan P, Allen J, et al. Ensembl 2021. Nucleic Acids Res 2021;49:D884-91. Crossref

21. Stelzer G, Rosen N, Plaschkes I, et al. The GeneCards suite: from gene data mining to disease genome sequence analyses. Curr Protoc Bioinformatics 2016;54:1.30.1-33. Crossref

22. GeneCards. Available from: www.genecards.org. Accessed 15 Jan 2024.

23. Rappaport N, Twik M, Plaschkes I, et al. MalaCards: an amalgamated human disease compendium with diverse clinical and genetic annotation and structured search. Nucleic Acids Res 2017;45:D877-87. Crossref

24. MalaCards. Available from: www.malacards.org. Accessed 15 Jan 2024.

25. WHO Working Group on the Clinical Characterisation and Management of COVID-19 infection. A minimal common outcome measure set for COVID-19 clinical research. Lancet Infect Dis 2020;20:e192-7. Crossref

26. Ulyanova V, Shah Mahmud R, Laikov A, et al. Anti-influenza activity of the ribonuclease binase: cellular targets detected by quantitative proteomics. Int J Mol Sci 2020;21:8294. Crossref

27. Hanson P, Zhang HM, Hemida MG, Ye X, Qiu Y, Yang D. IRES-dependent translational control during virus-induced endoplasmic reticulum stress and apoptosis. Front Microbiol 2012;3:92. Crossref

28. Miller MD, Farnet CM, Bushman FD. Human immunodeficiency virus type 1 preintegration complexes: studies of organization and composition. J Virol 1997;71:5382-90. Crossref

29. Frankel AD, Young JA. HIV-1: fifteen proteins and an RNA. Annu Rev Biochem 1998;67:1-25. Crossref

30. Kash JC, Goodman AG, Korth MJ, Katze MG. Hijacking of the host-cell response and translational control during influenza virus infection. Virus Res 2006;119:111-20. Crossref

31. van Oers MM, Doitsidou M, Thomas AA, de Maagd RA, Vlak JM. Translation of both 5’ TOP and non-TOP host mRNAs continues into the late phase of Baculovirus infection. Insect Mol Biol 2003;12:75-84. Crossref

32. Zhou H, Zhu J, Tu J, et al. Effect on virulence and pathogenicity of H5N1 influenza A virus through truncations of NS1 eIF4GI binding domain. J Infect Dis 2010;202:1338-46. Crossref

33. Foeger N, Kuehnel E, Cencic R, Skern T. The binding of foot-and-mouth disease virus leader proteinase to eIF4GI involves conserved ionic interactions. FEBS J 2005;272:2602-11. Crossref

34. Zaborowska I, Kellner K, Henry M, Meleady P, Walsh D. Recruitment of host translation initiation factor eIF4G by the Vaccinia Virus ssDNA-binding protein I3. Virology 2012;425:11-22. Crossref

35. Chartier-Harlin MC, Dachsel JC, Vilariño-Güell C, et al. Translation initiator EIF4G1 mutations in familial Parkinson disease. Am J Hum Genet 2011;89:398-406. Crossref

36. Bhoj EJ, Li D, Harr M, et al. Mutations in TBCK, encoding TBC1-domain-containing kinase, lead to a recognizable syndrome of intellectual disability and hypotonia. Am J Hum Genet 2016;98:782-8. Crossref

37. Chong JX, Caputo V, Phelps IG, et al. Recessive inactivating mutations in TBCK, encoding a Rab GTPase-activating protein, cause severe infantile syndromic encephalopathy. Am J Hum Genet 2016;98:772-81. Crossref

38. Khatoon F, Prasad K, Kumar V. COVID-19 associated nervous system manifestations. Sleep Med 2022;91:231-6. Crossref

39. Butowt R, von Bartheld CS. Anosmia in COVID-19: underlying mechanisms and assessment of an olfactory route to brain infection. Neuroscientist 2021;27:582-603. Crossref

40. Ye M, Ren Y, Lv T. Encephalitis as a clinical manifestation of COVID-19. Brain Behav Immun 2020;88:945-6. Crossref

41. Johansson A, Mohamed MS, Moulin TC, Schiöth HB. Neurological manifestations of COVID-19: a comprehensive literature review and discussion of mechanisms. J Neuroimmunol 2021;358:577658. Crossref

42. Paniz-Mondolfi A, Bryce C, Grimes Z, et al. Central nervous system involvement by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2). J Med Virol 2020;92:699-702. Crossref

43. Moriguchi T, Harii N, Goto J, et al. A first case of meningitis/encephalitis associated with SARS-coronavirus-2. Int J Infect Dis 2020;94:55-8. Crossref

44. von Weyhern CH, Kaufmann I, Neff F, Kremer M. Early evidence of pronounced brain involvement in fatal COVID-19 outcomes. Lancet 2020;395:e109. Crossref

45. Virhammar J, Kumlien E, Fällmar D, et al. Acute necrotizing encephalopathy with SARS-CoV-2 RNA confirmed in cerebrospinal fluid. Neurology 2020;95:445-9. Crossref

46. Radmanesh A, Derman A, Ishida K. COVID-19–associated delayed posthypoxic necrotizing leukoencephalopathy. J Neurol Sci 2020;415:116945. Crossref

47. Daamen AR, Bachali P, Owen KA, et al. Comprehensive transcriptomic analysis of COVID-19 blood, lung, and airway. Sci Rep 2021;11:7052. Crossref

48. Xiong Y, Liu Y, Cao L, et al. Transcriptomic characteristics of bronchoalveolar lavage fluid and peripheral blood mononuclear cells in COVID-19 patients. Emerg Microbes Infect 2020;9:761-70. Crossref

49. Gagliardi S, Poloni ET, Pandini C, et al. Detection of SARS-CoV-2 genome and whole transcriptome sequencing in frontal cortex of COVID-19 patients. Brain Behav Immun 2021;97:13-21. Crossref

50. Moni MA, Lin PI, Quinn JM, Eapen V. COVID-19 patient transcriptomic and genomic profiling reveals comorbidity interactions with psychiatric disorders. Transl Psychiatry 2021;11:160. Crossref

51. MAQC Consortium; Shi L, Reid LH, et al. The MicroArray Quality Control (MAQC) project shows inter- and intraplatform reproducibility of gene expression measurements. Nat Biotechnol 2006;24:1151-61. Crossref

Left ventricular thrombus (LVT) primarily occurs in patients who exhibit heart failure with reduced ejection fraction, particularly when these conditions are secondary to dilated cardiomyopathy or myocardial infarction. Recent advances in the treatment of myocardial ischaemia and heart failure have reduced the estimated incidence to 7 cases per 10 000 patients.1 However, this lower incidence does not reduce the importance of identifying and treating LVT; one study has shown very high risks of major cardiovascular or cerebrovascular events and mortality in patients with LVT.2

Although LVT has conventionally been managed with warfarin, multiple guidelines suggest different treatment algorithms based on expert opinion and small-scale studies, reflecting the lack of evidence that underlies such recommendations.3 4 This lack of evidence is partly related to the low incidence of LVT, which hinders adequately powered research with high evidence quality. Considering the growing popularity of non–vitamin K oral anticoagulants (NOACs), there has been increasing interest in the use of NOACs as an alternative to warfarin for LVT management.5 A systematic review in 2020, which involved only relevant case series and case reports, concluded that NOACs constitute a ‘reasonable alternative’ to warfarin for LVT management.6 However, another 2020 study of >500 patients showed that NOACs increased the incidence of stroke or systematic embolism compared with warfarin.7 Nonetheless, only thromboembolic events were compared in that study; safety outcomes, specifically bleeding events, were not investigated. Thus, it remains unclear whether NOACs exhibit efficacy and safety similar to warfarin for LVT management. This retrospective cohort study aimed to evaluate the efficacy and safety of NOACs versus warfarin for the treatment of LVT.

This retrospective cohort study included all patients with LVT diagnosed by echocardiography from January 2011 to January 2020 at our institution, a major tertiary university hospital in Hong Kong. Only patients aged ≥18 years were included. Patients were excluded if baseline echocardiography, pharmacotherapy regimen or clinical records were non-retrievable, or if the type of anticoagulation therapy (warfarin or NOACs) was switched within the first 2 years after LVT diagnosis.

At our institution, all patients began anticoagulation therapy upon echocardiography-based diagnosis of LVT. Patients either received warfarin with titration and maintenance of a therapeutic international normalised ratio of 2-3, or they received NOAC therapy. Because there are no specific treatment recommendations in current guidelines, anticoagulant selection was performed at the treating physicians’ discretion, generally considering patient-specific factors such as renal function, presence of other indications, and drug compliance. Follow-up echocardiography was performed 3 months after diagnosis of LVT, and further follow-up echocardiography was performed as clinically indicated. Anticoagulation was only discontinued if LVT had been resolved; this step required a shared, informed decision between the patient and the physician. Anticoagulation discontinuation was not considered for patients with persistent LVT.

All patients were followed up for ≤3 years. Echocardiographic images of all included patients at baseline and the 3-month follow-up were reviewed. The left ventricular ejection fraction, baseline size of LVT, and any resolution of LVT by the 3-month follow-up or the size of residual LVT at the 3-month follow-up were recorded. Clinical records of all patients were reviewed using the Clinical Management System of the Hong Kong Hospital Authority; important pre-morbid conditions, types of anticoagulants used, and pre-specified clinical outcomes were recorded. Any discontinuation of anticoagulation by 1 year was recorded.

The primary outcomes were cumulative mortality and net adverse clinical events (NACEs), defined as any of the following: ischaemic stroke, intracranial haemorrhage, systemic thromboembolism other than cerebral embolism, fatal bleeding (Bleeding Academic Research Consortium class 58), and major non-fatal bleeding (Bleeding Academic Research Consortium class 38). Secondary outcomes were complete resolution of LVT and percentage reduction of LVT size at the 3-month follow-up. Outcomes were also compared between patients who had discontinued anticoagulation by 1 year and those who continued anticoagulation for >1 year.

Unless otherwise specified, all continuous variables are expressed as mean ± standard deviation. Pre-morbid conditions and clinical outcomes in the two anticoagulation therapy groups were compared using Fisher’s exact test (for dichotomous variables) or Mann-Whitney U test (for continuous variables); the Mann-Whitney U test was chosen over parametric tests because the sample sizes were unlikely to support an assumption of data normality. Kaplan-Meier survival curves were used to visualise survival status and freedom from NACEs throughout the study period; Cox regression was used to compare mortality and NACE use between the two groups. Cases with missing values were excluded from analysis of the respective variables; no imputation was performed. All P values were two-sided, and P<0.05 was considered statistically significant. All statistical analyses were performed using SPSS software (Windows version 25.0; IBM Corp, Armonk [NY], United States).

In total, 43 patients (37 men) with LVT were included in this study: 28 received warfarin and 15 received NOACs. No patients were excluded for switching anticoagulant therapy during the first 2 years after LVT diagnosis. Of the patients treated with NOACs, 10 received apixaban, four received dabigatran, and one received rivaroxaban. Their baseline characteristics are summarised in Table 1; the two cohorts were generally comparable, except the NOAC cohort included more patients with diabetes mellitus (P=0.001) and atrial fibrillation or flutter (P=0.043). Eleven patients in the warfarin cohort and three patients in the NOAC cohort had non-ischaemic cardiomyopathy (P=0.308), including one patient with non-compaction cardiomyopathy and another patient (lost to follow-up after 6 months) with myocarditis. Both of these patients were in the warfarin cohort.

Three patients (all in the warfarin group) were lost to follow-up: one after 6 months (as noted above), one after 22 months, and one after 26 months. One of these patients had discontinued anticoagulation therapy by 1 year. The warfarin and NOAC cohorts were followed up for mean intervals of 20 ± 12 months (median, 20; interquartile range, 7-33) and 22 ± 9 months (median, 19; interquartile range, 15-31), respectively (P=0.522). All patients were examined by follow-up echocardiography at 3 months after initiation of anticoagulation therapy, except one patient in the warfarin cohort who died 1 month after diagnosis of LVT. In total, 14 deaths were observed in the NOAC (n=2; 13.3%) and warfarin (n=12; 42.9%) cohorts during the study period. Causes of death in the NOAC cohort were cardiovascular (sudden death; n=2); in the warfarin cohort, the causes of death were cardiovascular (n=8), intracerebral haemorrhage (n=3), gastrointestinal haemorrhage (n=1), and malignancy (n=2). Of the 34 patients who completed 1 year of follow-up, nine had discontinued anticoagulation therapy.

All primary and secondary outcomes are summarised in Table 2. We observed a significantly lower risk of NACEs in the NOAC cohort (n=1 [6.7%] in the NOAC cohort vs n=13 [46.4%] in the warfarin cohort; hazard ratio [HR]=0.124, 95% confidence interval [CI]=0.016-0.952; P=0.045), which remained statistically significant after adjustment for the clinical statuses of diabetes mellitus and atrial fibrillation or flutter (HR=0.111, 95% CI=0.012-0.994; P=0.049) [Fig 1]. There was a tendency towards lower mortality in the NOAC cohort (n=2 [13.3%] in the NOAC cohort vs n=12 [42.9%] in the warfarin cohort; HR=0.285, 95% CI=0.064-1.275; P=0.101 [before adjustment of clinical statuses]), which remained similar after adjustment for the clinical statuses of diabetes mellitus and atrial fibrillation or flutter (HR=0.184, 95% CI=0.032-1.059; P=0.058) [Fig 2]. Numerically lower rates of ischaemic stroke (n=0 [0%] in the NOAC cohort vs n=5 [17.9%] in the warfarin cohort), major non-fatal bleeding (n=0 [0%] in the NOAC cohort vs n=4 [14.3%] in the warfarin cohort), and fatal bleeding (n=1 [6.7%] in the NOAC cohort vs n=4 [14.3%] in the warfarin cohort) were observed among patients receiving NOACs.

Concerning secondary outcomes, there were no significant differences between the two cohorts in LVT resolution (P=0.451) or percentage reduction in LVT size (P=0.390) at the 3-month follow-up.

The outcomes of patients whose had or had not discontinued anticoagulation therapy by 1 year are summarised in Table 3. There were no significant differences between the two cohorts.

In this retrospective cohort study, we explored the use of NOACs as an alternative to warfarin for LVT management in a Hong Kong hospital. Although the sample size was limited, we found that NOAC use was associated with significantly fewer NACEs, with a tendency towards differences in cumulative survival. Additionally, anticoagulation discontinuation by 1-year post-diagnosis was not associated with significantly different clinical outcomes.

Our results confirm and extend previous findings concerning similar rates of LVT regression between NOAC and warfarin therapies; moreover, it has been reported that NOAC use is at least non-inferior to warfarin in terms of cumulative survival.2 Importantly, we demonstrated significantly lower rates of NACEs in NOAC users, a key finding that was likely driven by tendencies towards reductions in ischaemic stroke and major non-fatal bleeding. The numerically lower rate of major non-fatal bleeding in NOAC users was consistent with previous findings of lower bleeding risk among patients receiving NOACs compared with patients receiving warfarin.9 10 11 This reduction in bleeding risk is more prominent among Asian individuals than among non-Asian individuals.12 Therefore, it is possible that clinical practice recommendations for Asian individuals should be different from that for non-Asian individuals.

A recent study by Abdelnabi et al13 demonstrated significantly more effective resolution of LVT with rivaroxaban. We did not observe such difference, consistent with recent findings by Iqbal et al.14 These discrepancies may be related to differences in imaging intervals: we repeated echocardiography at 3 months and Iqbal et al14 repeated imaging at a mean interval of 233 days, whereas Abdelnabi et al13 repeated imaging at 1 month. Importantly, Abdelnabi et al13 observed converging rates of thrombus resolution by 3 and 6 months after initiation of anticoagulation, when they performed additional imaging. It is thus possible that frequent imaging intervals (more frequent than that recommended by societal guidelines3 4) are required to demonstrate differences in the rate of thrombus resolution. Although the clinical benefits of NOACs in our cohort were mainly driven by a reduction in bleeding events, more rapid thrombus resolution may be relevant in other populations. Further investigation in this area may be warranted.

Another recent study by Robinson et al7 revealed significantly higher rates of systemic thromboembolism among patients receiving NOACs, compared with those receiving warfarin. In the present study, systemic embolism was rare, and there were no pronounced numerical differences in the rates of systemic embolism between cohorts. Although this finding may be partly related to our small sample size, ethnic differences in thromboembolic tendencies could also play important roles. It has been observed that Asian individuals are generally less susceptible to thromboembolism than Caucasian and Hispanic individuals,15 consistent with the rarity of systemic thromboembolism in our cohort. These findings may imply that any increase in systemic thromboembolism associated with NOAC use, as detected by Robinson et al,7 is less relevant for Asian patients. Considering this lack of relevance and the reduction in NACEs observed in the present study, NOAC use may be preferrable to warfarin in Asian patients. Further studies with larger cohorts should be conducted to confirm these findings.

Additionally, we observed that considering the resolution of LVT, anticoagulation discontinuation by 1 year probably did not lead to significantly different rates of adverse outcomes, despite the numerically higher rate of cerebrovascular accidents. Although Lattuca et al2 showed that anticoagulation for ≥3 months reduced the incidence of major adverse cardiovascular events, it has been unclear whether anticoagulation can be discontinued after resolution of LVT. Our results, derived from a small cohort, warrant further investigation in larger cohorts.

There were several limitations in this study. First, the sample size was limited, primarily due to the rarity of LVT—although the study was conducted in a large tertiary hospital, only 43 patients could be included over a 9-year period. Second, various NOACs were used. Nonetheless, subgroup analysis was precluded by the small sample size; the present study design remains valid as a general comparison of vitamin K versus non–vitamin K anticoagulants, especially because all included NOACs are commonly prescribed. Third, more patients in the NOAC cohort had diabetes mellitus and atrial fibrillation or flutter. Despite these co-morbidities, we found that NOACs remained statistically superior to warfarin for NACEs; we also found a tendency for better cumulative mortality among patients receiving NOACs after adjustment for these two co-morbidities. Thus, our results remain valid in terms of demonstrating the probable superiority of NOACs over warfarin for LVT management in Asian patients.

The use of NOACs to treat patients with LVT was associated with significantly fewer NACEs, with a tendency towards lower cumulative mortality. Additionally, anticoagulation discontinuation by 1 year might be safe for patients with LVT resolution. Overall, NOACs may be superior to warfarin for LVT management. Further studies are required to confirm our findings and determine the optimal duration of anticoagulation therapy for LVT management.

Concept or design: KKH Kam, JSK Chan.
Acquisition of data: KKH Kam.
Analysis or interpretation of data: JSK Chan.
Drafting of the manuscript: JSK Chan.
Critical revision of the manuscript for important intellectual content: All authors.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

KKH Kam and JSK Chan have disclosed no conflicts of interest. APW Lee received grants, consulting fees/honoraria, and research support from Boehringer Ingelheim, Bayer, and Pfizer.

This research was presented as a poster at the European Society of Cardiology Congress 2021 (27-30 August 2021, online).

This research was supported by the Hong Kong Special Administrative Region Government Health and Medical Research Fund (Grant No.: 05160976). The funder had no role in study design, data collection/analysis/interpretation or manuscript preparation.

This research was approved by The Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee (Ref No.: 2020.425). The need for individual patient consent was waived by the Committee due to the retrospective nature of the study.

1. Lee JM, Park JJ, Jung HW, et al. Left ventricular thrombus and subsequent thromboembolism, comparison of anticoagulation, surgical removal, and antiplatelet agents. J Atheroscler Thromb. 2013;20:73-93. Crossref

2. Lattuca B, Bouziri N, Kerneis M, et al. Antithrombotic therapy for patients with left ventricular mural thrombus. J Am Coll Cardiol 2020;75:1676-85. Crossref

3. Ibanez B, James S, Agewall S, et al. 2017 ESC Guidelines for the management of acute myocardial infarction in patients presenting with ST-segment elevation: The Task Force for the management of acute myocardial infarction in patients presenting with ST-segment elevation of the European Society of Cardiology (ESC). Eur Heart J 2018;39:119-77. Crossref

4. Kernan WN, Ovbiagele B, Black HR, et al. Guidelines for the prevention of stroke in patients with stroke and transient ischemic attack: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke 2014;45:2160-236. Crossref

5. McCarthy CP, Murphy S, Venkateswaran RV, et al. Left ventricular thrombus: contemporary etiologies, treatment strategies, and outcomes. J Am Coll Cardiol 2019;73:2007-9. Crossref

6. Kajy M, Shokr M, Ramappa P. Use of direct oral anticoagulants in the treatment of left ventricular thrombus: systematic review of current literature. Am J Ther 2020;27:e584-90. Crossref

7. Robinson AA, Trankle CR, Eubanks G, et al. Off-label use of direct oral anticoagulants compared with warfarin for left ventricular thrombi. JAMA Cardiol 2020;5:685-92. Crossref

8. Mehran R, Rao SV, Bhatt DL, et al. Standardized bleeding definitions for cardiovascular clinical trials: a consensus report from the bleeding academic research consortium. Circulation 2011;123:2736-47. Crossref

9. Adeboyeje G, Sylwestrzak G, Barron JJ, et al. Major bleeding risk during anticoagulation with warfarin, dabigatran, apixaban, or rivaroxaban in patients with nonvalvular atrial fibrillation. J Manag Care Spec Pharm 2017;23:968-78. Crossref

10. Chan YH, See LC, Tu HT, et al. Efficacy and safety of apixaban, dabigatran, rivaroxaban, and warfarin in Asians with nonvalvular atrial fibrillation. J Am Heart Assoc 2018;7:e008150.Crossref

11. Patel P, Pandya J, Goldberg M. NOACs vs. warfarin for stroke prevention in nonvalvular atrial fibrillation. Cureus 2017;9:e1395. Crossref

12. Yamashita Y, Morimoto T, Toyota T, et al. Asian patients versus non-Asian patients in the efficacy and safety of direct oral anticoagulants relative to vitamin K antagonist for venous thromboembolism: a systemic review and meta-analysis. Thromb Res 2018;166:37-42. Crossref

13. Abdelnabi M, Saleh Y, Fareed A, et al. Comparative study of oral anticoagulation in left ventricular thrombi (No-LVT trial). J Am Coll Cardiol 2021;77:1590-2. Crossref

14. Iqbal H, Straw S, Craven TP, Stirling K, Wheatcroft SB, Witte KK. Direct oral anticoagulants compared to vitamin K antagonist for the management of left ventricular thrombus. ESC Hear Fail 2020;7:2032-41. Crossref

15. Zakai NA, McClure LA. Racial differences in venous thromboembolism. J Thromb Haemost 2011;9:1877-82. Crossref

Extracorporeal membrane oxygenation (ECMO) offers life-sustaining support by supplementing heart and lung functions in patients with circulatory or respiratory failure. There is increasing utilisation of ECMO in intensive care units (ICUs) worldwide; for example, the Extracorporeal Life Support Organization (ELSO) registry reported a 10-fold increase in ECMO runs from 1643 in 1990 to 18 260 in 2020.1 Hong Kong is a Special Administrative Region of the People’s Republic of China, with a population of 7.4 million and an independent healthcare system.2 In Hong Kong, various assessments of ICU performance have been performed for other disease entities,3 but there have been few reports of ECMO-specific data and patient outcomes.4 In particular, Hong Kong has a higher ECMO centre–to-population ratio compared with international guidelines.5 6 A retrospective study examined the risk score–mortality association in patients receiving ECMO, but it only included data from a single tertiary ICU and was not fully representative of territory-wide practices.7 Because ECMO is a high-cost, labour-intensive ICU treatment modality, it is important to understand how ECMO is utilised in Hong Kong, its associated resource implications, and review patient outcomes for future planning efforts.

In this study, using a territory-wide administrative registry of all patients receiving ECMO in the ICUs of public hospitals in Hong Kong, we examined trends in ECMO utilisation and clinical outcomes. Our primary objective was to summarise the status of ECMO services in Hong Kong over the past decade.

This retrospective observational study covered the period from 1 January 2010 to 31 December 2019. All adult patients aged ≥18 years with an ECMO episode and admission to the ICU of a public hospital under the Hospital Authority were identified using an administrative ECMO patient registry managed by a centralised ICU committee. An episode of ECMO was defined on the basis of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) procedure code for ECMO.8 The need for ICU admission was determined using the Acute Physiology and Chronic Health Evaluation IV (APACHE IV) evaluation form.9 10 Patients with missing ECMO details (eg, ECMO duration and configuration) and patients managed in non-mixed disciplinary ICUs were excluded from the study.

Extracorporeal membrane oxygenation data were extracted from the administrative patient registry, which contained information about ECMO configuration, time of initiation, and time of discontinuation that had been entered by qualified nurses at the corresponding ECMO centre. The Clinical Data Analysis and Reporting System (CDARS), a central de-identified data repository comprising electronic health records from all public hospitals in Hong Kong, was accessed to collect patient baseline characteristics and components of the following disease severity scores: Sequential Organ Failure Assessment (SOFA),11 Survival after Veno-Arterial ECMO (SAVE),12 Respiratory ECMO Survival Prediction (RESP),13 Charlson Comorbidity Index (CCI), and APACHE IV14 (online supplementary Table 1). For patients with multiple ICU admissions during a single hospital stay, the APACHE IV score for the first ICU admission was used. Clinical outcomes including length of stay (LOS) and mortality were retrieved from the CDARS.

The primary outcomes were trends in ECMO utilisation over 10 years, including number of patients receiving ECMO, illness severity (as measured by disease severity scores), and numbers of tertiary and quaternary inter-hospital transfers. Secondary outcomes were mortality, hospital and ICU LOS, transplantation procedures, ventricular assistive device (VAD) implantation, and complications. For patients who were transferred between hospitals, hospital mortality was defined as death during the final hospitalisation. Four common complications of ECMO, namely haemorrhagic, neurological, renal and cardiovascular complications, were identified using ICD-9-CM diagnostic and procedural codes (online supplementary Table 2).8 Major non-cranial bleeding was identified as the diagnosis of gastrointestinal, major internal, and/or postoperative bleeding; alternatively, it was identified by the need for haemostatic procedures, transfusion of >2 units of packed red blood cells over 24 hours, and/or use of recombinant factor VII. Stroke was subdivided into haemorrhagic and ischaemic types. Patients with acute ischaemic limbs were identified by the diagnosis of acute limb ischaemia or compartment syndrome or by the performance of limb-saving procedures (eg, fasciotomy and amputation). Brain death was identified by the appropriate diagnostic code or by a procedure code indicating organ collection from a deceased donor.

For the purposes of subsequent analyses, ECMO centres referred to designated ICUs under the governance of the Hospital Authority Central Organising Committee in ICU Services. An emergency admission was defined as an admission in which the patient had emergency room attendance records within the preceding 12 hours. Extracorporeal membrane oxygenation initiation in the emergency room was defined as an ECMO episode in which the patient had emergency room attendance records within the preceding 24 hours. An inter-hospital transfer was defined when ECMO was started at another institution before patient transferal with ECMO in situ to one of six ECMO centres. A transfer to a quaternary cardiothoracic unit was defined as an instance of intra- or inter-hospital transfer from a mixed ICU to cardiothoracic care in one of three centres, either during ECMO care or within 12 hours after stopping ECMO.

Frequencies and percentages were used to describe categorical variables. The Shapiro–Wilk test was used to assess data normality; data were expressed as means with standard deviations or medians with interquartile ranges, as appropriate. Categorical variables were compared between groups using the Chi squared test; continuous variables were compared by the t test or Mann-Whitney U test, as appropriate. The Mann–Kendall test was used to assess temporal trends in patient characteristics and outcomes across years, in sequential order from 2010 to 2019. Model discrimination and model calibration of risk scores in predicting hospital mortality were examined using the area under the receiver operating characteristic (AUROC) curve and the Hosmer–Lemeshow test. The observed hospital mortality was compared with that predicted from risk scores using standardised mortality ratios (SMRs). Patients with missing APACHE IV scores were excluded from this analysis.

All statistical analysis and data visualisation procedures were performed in Stata 16 (StataCorp; College Station [TX], United States). Tests were considered statistically significant when two-tailed P values were <0.05.

From January 2010 to December 2019, among 125 101 ICU admissions overall in Hong Kong, 911 (0.73%) involved patients receiving ECMO as follows: 297 (32.6%) veno-arterial (V-A) ECMO, 450 (49.4%) veno-venous (V-V) ECMO, and 164 (18.0%) extracorporeal cardiopulmonary resuscitation (ECPR) [Fig 1]. There was a steady increase in the annual number of patients receiving ECMO, with a 9.5-fold increase from 18 episodes in 2010 to 171 episodes in 2019 (Fig 2). The annual number of V-A ECMO episodes significantly increased from 3 (16.7%) to 67 (39.2%) over 10 years (P for trend=0.001) [Table 1]. The total number of ECMO patient-days increased from 109 in 2010 to 1565 in 2019 (online supplementary Fig 1).

A total of 583 (64.0%) patients were male, with a median age at admission of 54 years (interquartile range, 42-62), and 185 (20.3%) patients of ≥65 years. There was increasing utilisation of ECMO among patients aged ≥65 years (P for trend=0.001). The median CCI was 1 (0-2), and an increasing number of patients had a CCI ≥2 (P for trend=0.002) [Table 1].

Among the 889 (97.6%) patients with complete APACHE IV data, the median APACHE IV score was 100 (73-132), with an increase from 90 (57-112) in 2010 to 105 (77-137) in 2019 (P for trend=0.003); the median APACHE IV–estimated risk of death was 0.5 (0.2-0.8). Complete demographic details are shown in Table 1; trends in co-morbidities and disease severity scores are shown in online supplementary Figure 2.

Within the publicly funded hospital system, the number of ECMO centres under centralised ICU governance increased from three in 2010 to five in 2015, and then seven in 2019. The total number of available ECMO consoles paralleled the increase: from three in 2010 to nine in 2015, and then 11 in 2019 (online supplementary Table 3).

Among the 911 patients receiving ECMO, 469 (51.5%) were initiated outside of the regular 9 am to 5 pm period, including 247 (52.7%) patients receiving V-V ECMO, 137 (29.2%) patients receiving V-A ECMO, and 85 (18.1%) patients receiving ECPR. In total, 710 (77.9%) emergency admissions were identified. These patients were younger, had fewer co-morbidities, and were more likely to receive V-V ECMO [371/710 (52.3%) vs 79/201 (39.3%); P=0.001]. In total, 370 (40.6%) patients had ECMO initiated within 24 hours of emergency admission; these patients were more likely to receive ECPR [113/370 (30.5%) vs 51/541 (9.4%); P<0.001] and have higher APACHE IV scores [118 (86-146) vs 91 (69-118); P<0.001].

Overall, there were 222 (24.4%) episodes of inter-hospital transfer from non-ECMO centres to ECMO centres; the annual number of episodes increased from one (1/18 [5.6%]) in 2010 to 22 (22/171 [12.9%]) in 2019 (P for trend<0.001) [online supplementary Fig 3]. In total, 173 (77.9%) patients were transferred from ICUs in other hospitals; the remaining 49 patients were transferred from non-ICU settings. Most transferred patients (66.2%) received V-V ECMO (online supplementary Fig 4a); their principal diagnoses are shown in online supplementary Figure 4b and 4c. Transferred patients had worse RESP scores [-2 (-4 to 0) vs -1 (-3 to 2); P<0.001], better SAVE scores (-6 ± 5 vs -8 ± 5; P=0.012), and lower APACHE IV scores [89 (69-117) vs 104 (75-136); P<0.001]. Among the patients transferred to ECMO centres, 54 (24.3%) underwent a major operation within 7 days of transfer, and 32 (59.3%) of these surgeries involved the cardiovascular system. Other procedural details are shown in online supplementary Figure 4d and 4e.

There were 52 (5.7%) episodes of inter-hospital transfer to quaternary cardiothoracic ICUs; the annual number remained relatively consistent throughout the 10-year study period (P for trend=0.121) [online supplementary Fig 3]. Patients in these transfers were younger (P=0.048); they were more likely to receive V-A ECMO [31/52 (59.6%) vs 266/859 (31.0%); P<0.001] and ECPR [15/52 (28.8%) vs 149/859 (17.3%); P=0.036] (online supplementary Fig 5a). The primary diagnoses are shown in online supplementary Figure 5b and 5c. Among the patients involved in quaternary transfers, 22 (42.3%) underwent a major operation within 28 days of transfer, and 18 (81.8%) of these surgeries involved the cardiovascular system. Other procedural details are shown in online supplementary Figure 5d and 5e.

The overall numbers of hospital mortalities and ICU mortalities were 456 (50.1%) and 382 (41.9%), respectively. The numbers of hospital mortalities among patients receiving V-V ECMO, V-A ECMO, and ECPR were 152 (33.9%), 178 (59.9%), and 126 (76.8%), respectively (online supplementary Table 4). The median hospital LOS was 26.8 (interquartile range, 10.7-55.6) days, and the median ICU LOS was 10.2 (interquartile range, 4.8-20.1) days [Table 2]. Throughout the 10-year study period, the annual number of hospital mortalities increased from one (5.6%) in 2010 to 90 (52.6%) in 2019 (P for trend<0.001). The hospital LOS decreased from 36.6 (interquartile range, 26.8-57.2) to 25.2 (7.6-50.2) days [P for trend=0.011], and ICU LOS decreased from 15.5 (10.8-18.2) days in 2010 to 7.9 (3.9-19.8) days in 2019 (P for trend<0.001).

After adjustments for age, sex, APACHE IV score, and type of ECMO, the odds of hospital mortality were significantly lower in patients with ECMO initiated within 24 hours of emergency admission (adjusted odds ratio [OR]=0.56, 95% confidence interval [CI]=0.40-0.78; P=0.001). There were no significant associations with hospital mortality among patients who had emergency admission (adjusted OR=0.78, 95% CI=0.54- 1.12; P=0.17), patients who were transferred to ECMO centres (adjusted OR=0.74, 95% CI=0.52- 1.05; P=0.09), or patients who were transferred to quaternary cardiothoracic ICUs (adjusted OR=0.58, 95% CI=0.30-1.13; P=0.11). Patients transferred to quaternary cardiothoracic ICUs had significantly lower ICU mortality [4 (7.7%) vs 378 (44.0%); P<0.001] and significantly longer hospital LOS [38.3 (22.1-111.0) vs 25.6 (9.4-53.1) days, P<0.001]. The unadjusted and adjusted outcomes in various patient subgroups are presented in online supplementary Table 5.

In total, 41 (4.5%) patients were successfully bridged to VAD or transplantation. Among 461 patients who were receiving V-A ECMO and ECPR, 31 (6.7%) patients underwent VAD implantation and eight (1.7%) patients underwent heart transplantation. Among 450 patients who were receiving V-V ECMO, one (0.2%) patient underwent lung transplantation.

In terms of complications, there were 466 (51.2%) cases of major bleeding, 28 (3.1%) ischaemic limb complications, and nine (1.0%) patients who were declared brain-dead. Among 76 (8.3%) patients with stroke, 54 (5.9%) had haemorrhagic stroke (Table 2).

The ability of risk scores to predict post-ECMO hospital mortality was examined. There was a significant increase in the annual median APACHE IV score from 90 (57-112) in 2010 to 105 (77-137) in 2019 (P for trend=0.003). The SOFA score on the first day of ICU admission and the SAVE score in patients receiving V-A ECMO also showed significant trends (P for trend<0.001). No significant trends were observed regarding the SOFA score on the first day of ECMO (P for trend=0.58) or the RESP score in patients receiving V-V ECMO (P for trend=0.46) [Table 1].

The APACHE IV score showed good discriminatory power and was well calibrated for the prediction of hospital mortality (AUROC=0.727; Hosmer–Lemeshow test P=0.356); as was SOFA score on the first day of ECMO (AUROC=0.670; Hosmer–Lemeshow test P=0.322) [Fig 3]. The overall SMR for hospital mortality compared with APACHE IV–predicted mortality was 1.11 (95% CI=1.01-1.22) and there was no significant trend over the 10-year study period (P for trend=0.135) [Fig 4]. The SAVE and RESP scores, estimated using data from electronic health records, displayed limited discriminatory power for the prediction of hospital mortality in patients receiving V-A and V-V ECMO (AUROC=0.604 and 0.527, respectively). The ROC curves for various risk prediction models are shown in Fig 3.

To our knowledge, this is the first 10-year longitudinal study of the majority of patients receiving ECMO in Hong Kong; the results showed that the numbers of patients and complexities of medical conditions increased throughout the study period. Although patients receiving ECMO represent a small proportion of ICU patients overall, they require significant resource utilisation including out-of-hours services, inter-hospital transfers, and major operations. Comparisons with standardised risk scores suggested satisfactory performance based on the APACHE IV model, but the lack of complete and granular patient data precluded meaningful conclusions with respect to ECMO-specific risk scores.

In addition to the observation of a 9.5-fold increase in ECMO utilisation in Hong Kong over the 10-year study period, including greater use of V-A ECMO after 2012 and rapid uptake of ECPR after 2015, this study revealed that patients receiving ECMO were increasingly older, had an increasing co-morbidity burden, and displayed greater disease severity upon ICU admission. This is not only attributable to the overall advances in ECMO,15 but also encouraged by the multiple studies showing indistinguishable survival after ECMO in older adult patients compared with the younger ones.16 17 18 The increased utilisation of ECMO in ECPR is supported by clinical trials demonstrating the efficacy of this approach. In the CHEER trial (mechanical CPR, Hypothermia, ECMO and Early Reperfusion), treatment with mechanical cardiopulmonary resuscitation, hypothermia, ECMO, and early reperfusion led to increased survival among patients with refractory cardiac arrest.19 The ARREST trial (Advanced Reperfusion Strategies for Refractory Cardiac Arrest) showed a similar increase in survival upon initiation of early ECPR among patients with out-of-hospital cardiac arrest and refractory ventricular fibrillation.20 The increasing numbers of patients receiving ECMO have also resulted from greater utilisation of V-A ECMO to manage conditions such as acute myocardial infarction complicated by refractory cardiogenic shock,21 as well as efforts to transition to therapies including VAD and heart transplantation.22 The overall growth of ECMO utilisation in Hong Kong is similar to global patterns evident in the ELSO registry.1

The observed overall SMR for post-ECMO hospital mortality was slightly worse than the predicted overall SMR, possibly because ECMO services were in early phases of development at various centres throughout the study period. The decrease in SMR in the later portion of the study period, when ECMO services had matured at most centres, may be an indication of progress. When the results were stratified according to the type of ECMO, we found that the rate of hospital mortality among patients receiving V-V ECMO was better in Hong Kong than in the global ELSO registry (33.9% vs 40.8%), whereas the rates of hospital mortality among patients receiving V-A ECMO and ECPR were worse (V-A ECMO: 59.8% in Hong Kong vs 55.4% globally; ECPR: 76.8% in Hong Kong vs 69.8% globally). The sharp increase in ECPR utilisation may have contributed to an artificially elevated SMR, considering that ECPR is associated with worse survival relative to V-V ECMO and V-A ECMO19; notably, in a pilot cohort of patients receiving ECPR in Hong Kong, ICU survival was 32.4%.23 The low rate of ECMO bridging to transplantation in Hong Kong—nine (1.0%) patients over 10 years—also reduces overall cohort survival. Among developed countries/regions, Hong Kong has a very low rate of registration in the Centralised Organ Donation Register (3.8%) and limited motivation to participate in organ donation.24 Nevertheless, it remains important to actively explore methods to lower the SMR. One possibility involves consolidating ECMO services to a few specialised centres, based on evidence of a volume-outcome relationship repeatedly identified in other observational cohorts across various geographical regions and healthcare settings.25 26 Furthermore, a study in the United States showed that multidisciplinary interventions—including coordination among surgeons, cardiologists, and ECMO specialists, as well as the implementation of standardised ECMO admission and weaning protocols—were associated with lower mortality in patients receiving ECMO,27 indicating a need to strengthen interdisciplinary communication or expand collaborations with allied health services to maintain standards of care.

The comparative utility of various risk scores for outcome prediction in Hong Kong merits attention. In terms of predicting hospital mortality among patients receiving ECMO in the present study, the APACHE IV score performed best, followed by the SOFA score on the first day of ECMO; the SAVE and RESP scores had moderate discriminatory power. The satisfactory performance of the APACHE IV score in Hong Kong was previously demonstrated in a large retrospective cohort study of ICU patients (c-statistic=0.889).28 Importantly, most data were available for APACHE IV scores in the present study, and the corresponding accuracy was high. However, the main limitation of APACHE IV scores is the lack of definite correlation with the time and patient condition upon ECMO initiation,29 which likely leads to a systemic under-representation of disease severity. The SOFA score, which can be calculated on a daily basis, has the theoretical advantage of more closely reflecting disease severity and clinical progression30; the SOFA score on the date of ECMO initiation demonstrated good performance in predicting hospital mortality among patients in our cohort. We note that the limited predictive performances of ECMO-specific SAVE and RESP scores are mainly related to the difficulty of retrieving accurate physiological data from the CDARS; various components of the scores were determined by a combination of diagnostic codes, procedural codes, and laboratory parameters. Although these scores have been validated in international cohorts,12 13 their systematic adoption as benchmarks for ECMO service performance in Hong Kong is hindered by the lack of available patient data. Among the six ECMO centres included in the present study, only four routinely collect patient and ECMO data for submission to the international ELSO registry; none of the centres compute SAVE and RESP scores. Within the community of ECMO providers in Hong Kong, we strongly encourage collaborative efforts to routinely document ECMO-specific severity scores and improve coding practices within electronic health records and the CDARS; these approaches will facilitate outcome monitoring and resource allocation. Moreover, validation of these scores in Hong Kong will be informative because Asians were substantially underrepresented in the original scoredevelopment cohorts established using the ELSO international registry.12 13

There were some limitations in this study. First, the retrospective observational design utilised data that were not recorded in a manner intended for research purposes; systematic biases in missing data may be present. Inaccurate diagnoses and procedural coding practices may have led to insufficient collection of relevant clinical data and information regarding ECMO circuit complications. However, the clinical outcomes of hospital mortality and LOS were captured from administrative data with a low risk of error. Second, the presence of between-centre heterogeneity related to non-uniform clinical practices may have contributed to outcome differences that were not reflected in the overall cohort. Third, patients receiving ECMO in non-mixed disciplinary ICUs or coronary care units were excluded from the study; outcomes and resource utilisation may have been considerably different among these patients. Finally, the collected data did not allow examination of ECMO cost-effectiveness, an important metric for service and resource planning.

In this territory-wide study, we observed increasing trends in ECMO utilisation in Hong Kong that were similar to global patterns. The overall observed mortality was reasonably close to the APACHE IV–predicted mortality. Systematic documentation of ECMO-specific risk scores is needed to ensure high-quality data for ECMO service benchmarking and development efforts.

Concept or design: PY Ng, A Ip.
Acquisition of data: VWS Chan, A Ip.
Analysis or interpretation of data: PY Ng, VWS Chan.
Drafting of the manuscript: PY Ng, VWS Chan.
Critical revision of the manuscript for important intellectual content: All authors.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

All authors have disclosed no conflicts of interest.

This research was supported by an unrestricted philanthropic donation from Mr and Mrs Laurence Tse. The funder had no role in study design, data collection, analysis and interpretation of the data, or manuscript preparation.

This research was approved by the Institutional Review Board of The University of Hong Kong/Hospital Authority Hong Kong West Cluster (Ref No.: UW 20-573). The research was conducted in accordance with the Declaration of Helsinki. The requirement for informed consent was waived by the Board due to the retrospective nature of the research.

1. Extracorporeal Life Support Organization. ECLS Registry Report. International Summary. October 2021. Report data through 2020. Available from: https://www.elso.org/Portals/0/Files/Reports/2021_October/International%20Report%20October_page1.pdf. Accessed 20 Dec 2021.

2. Department of Health, Hong Kong SAR Government. Health Facts of Hong Kong. 2021 Edition. Available from: https://www.dh.gov.hk/english/statistics/statistics_hs/files/2021.pdf. Accessed 20 Dec 2021

3. Lam KW, Lai KY. Evaluation of outcome and performance of an intensive care unit in Hong Kong by APACHE IV model: 2007-2014. J Emerg Crit Care Med 2017;1:16. Crossref

4. Sin SW, Young K. Development of extracorporeal membrane oxygenation in Hong Kong: current challenges and future development. Hong Kong Med J 2017;23:216-7. Crossref

5. Combes A, Brodie D, Bartlett R, et al. Position paper for the organization of extracorporeal membrane oxygenation programs for acute respiratory failure in adult patients. Am J Respir Crit Care Med 2014;190:488-96. Crossref

6. Ng PY, Ip A, Fang S, et al. Effect of hospital case volume on clinical outcomes of patients requiring extracorporeal membrane oxygenation: a territory-wide longitudinal observational study. J Thorac Dis 2022;14:1802-14. Crossref

7. Ng WT, Ling L, Joynt GM, Chan KM. An audit of mortality by using ECMO specific scores and APACHE II scoring system in patients receiving extracorporeal membrane oxygenation in a tertiary intensive care unit in Hong Kong. J Thorac Dis 2019;11:445-55. Crossref

8. National Center for Health Statistics, Centers for Disease Control and Prevention. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Available from: https://www.cdc.gov/nchs/icd/icd9cm.htm. Accessed 10 Oct 2023.

9. Zimmerman JE, Kramer AA, McNair DS, Malila FM. Acute Physiology and Chronic Health Evaluation (APACHE) IV: hospital mortality assessment for today's critically ill patients. Crit Care Med 2006;34:1297-310. Crossref

10. Zimmerman JE, Kramer AA, McNair DS, Malila FM, Shaffer VL. Intensive care unit length of stay: benchmarking based on Acute Physiology and Chronic Health Evaluation (APACHE) IV. Crit Care Med 2006;34:2517-29. Crossref

11. Vincent JL, Moreno R, Takala J, et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. Intensive Care Med 1996;22:707-10. Crossref

12. Schmidt M, Burrell A, Roberts L, et al. Predicting survival after ECMO for refractory cardiogenic shock: the survival after veno-arterial-ECMO (SAVE)–score. Eur Heart J 2015;36:2246-56. Crossref

13. Schmidt M, Bailey M, Sheldrake J, et al. Predicting survival after extracorporeal membrane oxygenation for severe acute respiratory failure. The Respiratory Extracorporeal Membrane Oxygenation Survival Prediction (RESP) score. Am J Respir Crit Care Med 2014;189:1374-82. Crossref

14. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373-83. Crossref

15. Mosier JM, Kelsey M, Raz Y, et al. Extracorporeal membrane oxygenation (ECMO) for critically ill adults in the emergency department: history, current applications, and future directions. Crit Care 2015;19:431. Crossref

16. Lee SN, Jo MS, Yoo KD. Impact of age on extracorporeal membrane oxygenation survival of patients with cardiac failure. Clin Interv Aging 2017;12:1347-53. Crossref

17. Narotsky DL, Mosca MS, Mochari-Greenberger H, et al. Short-term and longer-term survival after veno-arterial extracorporeal membrane oxygenation in an adult patient population: does older age matter? Perfusion 2016;31:366-75. Crossref

18. Saito S, Nakatani T, Kobayashi J, et al. Is extracorporeal life support contraindicated in elderly patients? Ann Thorac Surg 2007;83:140-5. Crossref

19. Stub D, Bernard S, Pellegrino V, et al. Refractory cardiac arrest treated with mechanical CPR, hypothermia, ECMO and early reperfusion (the CHEER trial). Resuscitation 2015;86:88-94. Crossref

20. Yannopoulos D, Bartos J, Raveendran G, et al. Advanced reperfusion strategies for patients with out-of-hospital cardiac arrest and refractory ventricular fibrillation (ARREST): a phase 2, single centre, open-label, randomised controlled trial. Lancet 2020;396:1807-16. Crossref

21. Tsao NW, Shih CM, Yeh JS, et al. Extracorporeal membrane oxygenation–assisted primary percutaneous coronary intervention may improve survival of patients with acute myocardial infarction complicated by profound cardiogenic shock. J Crit Care 2012;27:530.e1-11. Crossref

22. Brugts JJ, Caliskan K. Short-term mechanical circulatory support by veno-arterial extracorporeal membrane oxygenation in the management of cardiogenic shock and end-stage heart failure. Expert Rev Cardiovasc Ther 2014;12:145-53. Crossref

23. Ng PY, Li AC, Fang S, et al. Predictors of favorable neurologic outcomes in a territory-first extracorporeal cardiopulmonary resuscitation program. ASAIO J 2022;68:1158-64. Crossref

24. Tsai NW, Leung YM, Ng PY, et al. Attitudes of visitors at adult intensive care unit toward organ donation and organ support. Chin Med J (Engl) 2019;132:373-6. Crossref

25. Muguruma K, Kunisawa S, Fushimi K, Imanaka Y. Epidemiology and volume-outcome relationship of extracorporeal membrane oxygenation for respiratory failure in Japan: a retrospective observational study using a national administrative database. Acute Med Surg 2020;7:e486. Crossref

26. Barbaro RP, Odetola FO, Kidwell KM, et al. Association of hospital-level volume of extracorporeal membrane oxygenation cases and mortality. Analysis of the extracorporeal life support organization registry. Am J Respir Crit Care Med 2015;191:894-901. Crossref

27. Ratnani I, Tuazon D, Zainab A, Uddin F. The role and impact of extracorporeal membrane oxygenation in critical care. Methodist Debakey Cardiovasc J 2018;14:110-9. Crossref

28. Ling L, Ho CM, Ng PY, et al. Characteristics and outcomes of patients admitted to adult intensive care units in Hong Kong: a population retrospective cohort study from 2008 to 2018. J Intensive Care 2021;9:2. Crossref

29. Ko M, Shim M, Lee SM, Kim Y, Yoon S. Performance of APACHE IV in medical intensive care unit patients: comparisons with APACHE II, SAPS 3, and MPM0 III. Acute Crit Care 2018;33:216-21. Crossref

30. Lambden S, Laterre PF, Levy MM, Francois B. The SOFA score–development, utility and challenges of accurate assessment in clinical trials. Crit Care 2019;23:374. Crossref

Miscarriage occurs in 10% to 15% of pregnancies, mainly in the first trimester.1 Spontaneous miscarriage is associated with psychological problems such as anxiety and depression.2 3 4 Post-traumatic stress disorder may also occur after a miscarriage.3 Threatened miscarriage affects 15% to 20% of pregnant women.5 Management of threatened miscarriage involves reassurance and counselling. Women with threatened miscarriage and/or pregnancy-related uncertainty may experience frustration and anxiety. Although there are extensive data regarding the association between miscarriage and psychological morbidity, higher incidences of anxiety and depression among women with threatened miscarriage have only been detected in small studies.6 7

Antenatal depression and anxiety disorders are associated with increased fetal risks, such as low birth weight; antenatal symptoms of depression have been positively associated with postnatal depression.8 9 Furthermore, antenatal maternal stress and anxiety appear to predict long-term behavioural and emotional problems in children.10 11 Therefore, early detection and intervention are needed in women with antenatal psychological symptoms to minimise the impacts of those symptoms.

This study assessed psychological morbidity in women with threatened miscarriage, with the goal of identifying early interventions for women at risk of anxiety or depression.

This cross-sectional study, conducted at a university hospital in Hong Kong between July 2013 and June 2015, was part of a study that examined the ability of an early pregnancy viability scoring system to support counselling for women.12 In this hospital, an outpatient Early Pregnancy Assessment Clinic (EPAC) provides medical care for women experiencing abdominal pain, vaginal bleeding, or other problems in early pregnancy (gestational age ≤12 weeks). Referrals are usually made by medical officers from the Accident and Emergency Departments across Hong Kong, as well as general practitioners. All gynaecologists at the EPAC have ≥3 years of experience performing ultrasound scans. All Chinese women attending the EPAC were invited to participate; written informed consent was obtained from women who agreed to take part in the study.

Exclusion criteria were age <18 years, ectopic pregnancy, multiple pregnancies, gestational age >84 days (>12 weeks), requested termination of pregnancy, and loss to follow-up. Demographic data, obstetric history, smoking status, alcohol consumption, and body mass index were recorded. Details of early pregnancy complaints were assessed, including abdominal pain (graded by a pain score of 0 to 3, where a higher score represents greater pain) and vaginal bleeding (determined by a pictorial blood loss chart according to the number of pads used, where 0 corresponds to no bleeding and 4 corresponds to clots or flooding).

Chinese validated versions of five questionnaires were used to assess psychological well-being among the participants: the 12-item General Health Questionnaire (GHQ-12), the Beck Depression Inventory (BDI), Spielberger’s State Anxiety Inventory State form (STAI-S), the Fatigue Scale–14 (FS-14), and the Profile of Mood States (POMS). All questionnaires have demonstrated reliability and validity in previous studies.13 14 15 The GHQ-12 and BDI scores were reported as both continuous and categorical variables, while the scores of other questionnaires were reported as continuous variables.

The GHQ-12 is a self-reporting rating scale intended to identify individuals with reduced psychological well-being or diminished quality of life. It is sensitive to short-term psychiatric disorders. A total score of ≥4 using the bi-modal scoring method (0-0-1-1) is considered ‘high distress’. When using the Likert scoring method (0-1-2-3), scores ≤12 are considered normal, while scores >12 are considered evidence of psychological distress.16 17 The questionnaire has been used as a tool to evaluate women with miscarriage.18 19 20

The BDI is a 21-item self-reporting rating scale intended to measure symptoms of depression in both general and psychiatric populations.21 It is used to measure depression severity, and psychological morbidity is defined as a score of >12, indicating probable depressive disorder. The STAI-S is a 20-item self-reporting inventory that measures state anxiety, including transitory and situational feelings of worry.22 Its use has been validated in pregnant women.23 A higher score indicates a higher level of anxiety. The FS-14 is a 14-item self-rating questionnaire that measures fatigue severity. A lower score indicates a higher level of fatigue.

The POMS is a self-reporting tool for the assessment of mood alterations in clinical and psychiatric populations.24 This 65-item questionnaire contains seven components: tension-anxiety, depression-dejection, anger-hostility, fatigue-inertia, confusion-bewilderment, vigour-activity, and total mood disturbance. Scores range from 0 (not at all) to 4 (extremely). Higher positive mood scores indicate an ideal mood, whereas higher negative mood scores indicate severe mood disturbance.

In addition to the above questionnaires, each woman’s level of anxiety and worry before consultation was assessed using a 0 to 10 cm visual analogue scale (VAS). A higher value represents a higher level of anxiety and worry about their pregnancies. The VAS has previously been validated with respect to its correlations with other measures of anxiety.25 26

During the consultation, women were counselled based on their ultrasound findings and clinical diagnosis. These women used the VAS to indicate their level of pregnancy-related anxiety after consultation. Follow-up scans (1-2 weeks later) were offered for women with a pregnancy of uncertain viability. Actual pregnancy outcomes were reassessed at 13 to 16 weeks, either by phone or by retrieval of information from the hospital’s centralised computer antenatal records system.

The sample size was calculated based on the number of participants required for the primary study to validate the scoring system.12 SPSS software (Windows version 26.0; IBM Corp, Armonk [NY], United States) was used for data entry and analysis. A 95% confidence interval (95% CI) was calculated to determine the estimated errors and prevalence. Descriptive analyses were used for demographic data. The Chi squared test was used to explore associations between categorical variables. The Mann-Whitney U test was used to compare median values when data were not normally distributed, while the t test was used to compare means when data were normally distributed. Univariate analyses were performed to identify factors associated with psychological distress or morbidity. Factors with P values <0.1 in univariate analysis were entered into multivariate analysis, which was conducted via binary logistic regression. P values <0.05 were considered statistically significant.

Among the 1508 women who attended the EPAC during the study period, 64 were excluded and 54 declined to participate; thus, 1390 women completed the study (Fig). The demographic data are shown in Table 1. At the first clinic visit, most women (n=1048, 75.4%) had a viable pregnancy, 223 women (16.0%) had a pregnancy of uncertain viability, and 119 (8.6%) women had a miscarriage. At 13 to 16 weeks of gestation, 1111 women (79.9%) had a viable pregnancy, and an additional 160 (11.5%) women had a miscarriage.

The GHQ-12 (both bi-modal and Likert), BDI, STAI-S, FS-14, POMS, and VAS results are presented in Table 2. Overall, 48.4% and 76.7% of women had a GHQ-12 (bi-modal) score ≥4 and a GHQ-12 (Likert) score >12, respectively, indicating distress. Among the viable pregnancy, uncertain viability, and miscarriage groups, the percentages of women with a GHQ-12 (bi-modal) score ≥4 (43-52%) and a GHQ-12 (Likert) score >12 (73-83%) were similar. Women with miscarriage had the highest GHQ-12 score and the highest percentages of a GHQ-12 (bi-modal) score ≥4 and a GHQ-12 (Likert) score >12. The miscarriage group also had relatively higher POMS subscale scores for tension-anxiety, depression-dejection, anger-hostility, confusion-bewilderment, and total mood disturbance compared with women who had other diagnoses.

The VAS scores for anxiety are also presented in Table 2. The score before consultation was the highest among women with a miscarriage (mean ± standard deviation=7.02 ± 2.50). Although the scores for the viable pregnancy and uncertain viability groups were significantly lower after consultation, the score was substantially higher in the uncertain viability group than in the viable pregnancy group.

Subgroup analysis showed that GHQ-12, BDI, and POMS (except fatigue-inertia and vigour-activity subscales) scores were significantly higher among women with previous miscarriage than among those without (Table 3). The bleeding score was strongly positively correlated with the GHQ-12 (Likert) score (correlation coefficient=0.56; P=0.032). Univariate analysis revealed that compared with women who had a lower bleeding score (<2), women with a higher bleeding score (≥2) had a significantly higher risk of having a GHQ-12 (bi-modal) score ≥4 (P=0.041), a GHQ-12 (Likert) score >12 (P=0.025), and a BDI score >12 (P=0.022) [Table 4]. There were statistically significant but weak positive correlations between the pain score and a GHQ-12 (bi-modal) score ≥4 (P=0.001), a BDI score >12 (P<0.001), and a POMS total mood disturbance score (P<0.001), as well as various subscales. Notably, the POMS confusion-bewilderment subscale (correlation coefficient=0.93; P<0.001) demonstrated a strong positive correlation with the pain score.

Statistically significant factors associated with the various psychometric instrument scores were subjected to multivariate analysis (Table 4). Previous miscarriage was an independent risk factor for a GHQ-12 (bi-modal) score ≥4 (odds ratio [OR]=1.570, 95% CI=1.19-2.07) and a GHQ-12 (Likert) score >12 (OR=1.459, 95% CI=1.04-2.06), indicating distress; it was also a risk factor for a BDI score >12 (OR=1.717, 95% CI=1.28-2.30), suggesting probable depression. A bleeding score ≥2 was an independent risk factor for a GHQ-12 (Likert) score >12 (OR=1.506, 95% CI=1.05-2.17) and a BDI score >12 (OR=1.423, 95% CI=1.04-1.95).

In this cohort study, nearly 50% and approximately 77% of women had a GHQ-12 (bi-modal) score ≥4 and a GHQ-12 (Likert) score >12, indicating distress. Around one-fourth (24.5%) of women had a BDI score >12, suggesting probable depression.

Regardless of diagnosis, the VAS score decreased after consultation. However, the decrease in VAS score was smaller for the uncertain viability group, which may be attributed to the enhanced anxiety resulting from uncertainty among these women. This anxiety would be alleviated after an ultrasound examination and consultation with an accurate diagnosis. These findings emphasise the need to implement an early pregnancy assessment service that provides both clinical and psychological guidance to alleviate anxiety among women with problems in early pregnancy.

Emotional disturbances can have long-term effects on women with a previous miscarriage. Lok et al19 reported consistently higher scores on the GHQ-12 and BDI among women with a previous miscarriage, although these scores could decrease over time. In the present study, we observed a higher level of distress among women with a previous miscarriage, as demonstrated by the significantly greater proportion of women with GHQ-12 (bi-modal) score ≥4, GHQ-12 (Likert) score >12, and BDI score >12. Profile of Mood States scores were also significantly higher on all subscales, except for the fatigue-inertia and vigour-activity subscales. Similarly, the baseline VAS score before consultation was significantly higher among women with a previous miscarriage than among those without. In multivariate analysis, previous miscarriage was an independent risk factor for GHQ-12 (bi-modal) score >4, GHQ-12 (Likert) score >12, and BDI score >12. Baseline psychological morbidity may be greater among women with a previous miscarriage than among those without, consistent with findings in other studies.3 27 Therefore, additional attention and psychological support would be beneficial for women with greater distress and worse mood status.

A higher pain score was positively correlated with higher levels of distress and anxiety, as indicated by the positive relationships with various scales used in the present study. Pain is associated with anxiety and depression in pregnant women.28 Nevertheless, we observed weak relationships between pain and anxiety or distress, which might be related to the subjective nature of pain assessment.

Women with moderate to heavy bleeding (bleeding score ≥2) had significantly higher GHQ-12 (bi-modal), GHQ-12 (Likert), and BDI scores. Additionally, multivariate analysis showed that moderate to heavy bleeding (bleeding score ≥2) was an independent risk factor for a GHQ-12 (bi-modal) score ≥4, a GHQ-12 (Likert) score >12, and a BDI score >12. Heavy bleeding is often regarded as a common sign of threatened miscarriage. These findings highlight the importance of addressing pain and bleeding symptoms among women who attend early pregnancy services. The underlying complications of pregnancy, as well as anxiety and low mood in affected women, should be promptly managed.

Pregnancy loss is associated with negative mood status, including depression and anxiety.3 18 19 27 29 Whereas many studies have investigated the effects of miscarriage or pregnancy loss on depression, the effects of threatened miscarriage or early pregnancy-related complaints on women have not been extensively explored, despite the burdensome experience of a threatened miscarriage that appropriately causing anxiety in affected women. Our results are consistent with findings by Zhu et al,6 who reported that a substantial proportion of women with threatened miscarriage had symptoms of depression or anxiety.

The present study had a large sample size and a high rate of participation. Additionally, multiple psychometric instruments were used to assess the participants. The findings emphasise the importance of assessing and managing depression and anxiety symptoms in women with threatened miscarriage. Mental health assessments should be performed when women with threatened miscarriage attend clinics and hospitals. Early recognition of relevant mood problems will facilitate timely management. Non-pharmacological interventions, such as antenatal group therapy, constitute effective treatment for pregnant women with anxiety and depression.6 Pharmacological therapies (eg, most selective serotonin reuptake inhibitors and benzodiazepines) can be administered after considering the side-effects of medications relative to the risk of untreated antenatal depression and anxiety.30

Nevertheless, this study had some limitations. First, it used a cross-sectional design without longitudinal follow-up, and the subgroup analysis might have been underpowered. Second, information was unavailable regarding social factors (eg, education level or marital status) and the presence of an underlying psychiatric disorder, which might contribute to differences in baseline mood status. Third, the study did not include a comparison group of women without symptoms of threatened miscarriage.

There is a considerable psychological burden among women with early pregnancy problems and concerns about future pregnancy viability. These women experience emotional disturbances, as indicated by a significant proportion of women in this study who had high scores on psychometric tests. A gynaecologist consultation, in combination with an ultrasound assessment, is reassuring and can alleviate anxiety among women with early pregnancy problems. This study on maternal psychological outcomes provides insights concerning psychological morbidity among women with threatened miscarriage in the first trimester, while also demonstrating the usefulness and feasibility of various psychometric instruments in identifying women who require additional psychological support. Further studies exploring maternal psychological well-being later in pregnancy, as well as fetal outcomes, are needed to determine the long-term effects of anxiety and depression among women with threatened miscarriage in the first trimester.

Concept or design: OYK Wan, SSC Chan.
Acquisition of data: OYK Wan, JWK Kwok.
Analysis or interpretation of data: PNP Ip, K Ng.
Drafting of the manuscript: PNP Ip, K Ng.
Critical revision of the manuscript for important intellectual content: PNP Ip, K Ng, OYK Wan, JPW Chung, SSC Chan.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

As an editor of the journal, JPW Chung was not involved in the peer review process. Other authors have disclosed no conflicts of interest.

This research was supported by a grant from the Health and Medical Research Fund of the former Food and Health Bureau, Hong Kong SAR Government (Ref No.: 12131091). The study sponsor was not involved in the collection, analysis, or interpretation of data, or in the writing of the manuscript.

This research was approved by the Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee (Ref No.: CRE.2013.348). Written informed consent was obtained from all participants.

1. Tong S, Kaur A, Walker SP, Bryant V, Onwude JL, Permezel M. Miscarriage risk for asymptomatic women after a normal first-trimester prenatal visit. Obstet Gynecol 2008;111:710-4. Crossref

2. Broen AN, Moum T, Bödtker AS, Ekeberg O. Predictors of anxiety and depression following pregnancy termination: a longitudinal five-year follow-up study. Acta Obstet Gynecol Scand 2006;85:317-23. Crossref

3. Farren J, Jalmbrant M, Ameye L, et al. Post-traumatic stress, anxiety and depression following miscarriage or ectopic pregnancy: a prospective cohort study. BMJ Open 2016;6:e011864. Crossref

4. Jensen KL, Temple-Smith MJ, Bilardi JE. Health professionals’ roles and practices in supporting women experiencing miscarriage: a qualitative study. Aust N Z J Obstet Gynaecol 2019;59:508-13. Crossref

5. Park C, Kang MY, Kim D, Park J, Eom H, Kim EA. Prevalence of abortion and adverse pregnancy outcomes among working women in Korea: a cross-sectional study. PLoS One 2017;12:e0182341. Crossref

6. Zhu CS, Tan TC, Chen HY, Malhotra R, Allen JC, Østbye T. Threatened miscarriage and depressive and anxiety symptoms among women and partners in early pregnancy. J Affect Disord 2018;237:1-9. Crossref

7. Aksoy H, Aksoy Ü, İdem Karadağ Ö, et al. Effect of threatened miscarriage on maternal mood: a prospective controlled cohort study. J Clin Obstet Gynecol 2015;25:92-8. Crossref

8. Gelaye B, Sanchez SE, Andrade A, et al. Association of antepartum depression, generalized anxiety, and posttraumatic stress disorder with infant birth weight and gestational age at delivery. J Affect Disord 2020;262:310-6. Crossref

9. Siu BW, Leung SS, Ip P, Hung SF, O’Hara MW. Antenatal risk factors for postnatal depression: a prospective study of Chinese women at maternal and child health centres. BMC Psychiatry 2012;12:22. Crossref

10. Ibanez G, Bernard JY, Rondet C, et al. Effects of antenatal maternal depression and anxiety on children’s early cognitive development: a prospective cohort study. PLoS One 2015;10:e0135849. Crossref

11. O’Connor TG, Heron J, Golding J, Glover V; ALSPAC Study Team. Maternal antenatal anxiety and behavioural/emotional problems in children: a test of a programming hypothesis. J Child Psychol Psychiatry 2003;44:1025-36. Crossref

12. Wan OY, Chan SS, Chung JP, Kwok JW, Lao TT, Sahota DS. External validation of a simple scoring system to predict pregnancy viability in women presenting to an early pregnancy assessment clinic. Hong Kong Med J 2020;26:102-10. Crossref

13. Chan DW, Chan TS. Reliability, validity and the structure of the General Health Questionnaire in a Chinese context. Psychol Med 1983;13:363-71. Crossref

14. Shek DT. Reliability and factorial structure of the Chinese version of the Beck Depression Inventory. J Clin Psychol 1990;46:35-43. Crossref

15. Shek DT. Reliability and factorial structure of the Chinese version of the State-Trait Anxiety Inventory. J Psychopathol Behav Assess 1988;10:303-17. Crossref

16. Liang Y, Wang L, Yin X. The factor structure of the 12-item General Health Questionnaire (GHQ-12) in young Chinese civil servants. Health Qual Life Outcomes 2016;14:136. Crossref

17. Rao WW, Yang MJ, Cao BN, et al. Psychological distress in cancer patients in a large Chinese cross-sectional study. J Affect Disord 2019;245:950-6. Crossref

18. Lok IH, Lee DT, Yip SK, Shek D, Tam WH, Chung TK. Screening for post-miscarriage psychiatric morbidity. Am J Obstet Gynecol 2004;191:546-50. Crossref

19. Lok IH, Yip AS, Lee DT, Sahota D, Chung TK. A 1-year longitudinal study of psychological morbidity after miscarriage. Fertil Steril 2010;93:1966-75. Crossref

20. Kong GW, Lok IH, Yiu AK, Hui AS, Lai BP, Chung TK. Clinical and psychological impact after surgical, medical or expectant management of first-trimester miscarriage—a randomised controlled trial. Aust N Z J Obstet Gynaecol 2013;53:170-7. Crossref

21. Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. An inventory for measuring depression. Arch Gen Psychiatry 1961;4:561-71. Crossref

22. Hedberg AG. Review of State–Trait Anxiety Inventory. [Review of the Book State–Trait Anxiety Inventory, by C. D. Spielberger, R. L. Gorsuch & R. E. Lushere]. Prof Psychol 1972;3:389-90. Crossref

23. Bann CM, Parker CB, Grobman WA, et al. Psychometric properties of stress and anxiety measures among nulliparous women. J Psychosom Obstet Gynaecol 2017;38:53-62. Crossref

24. McNair DM, Heuchert P. Profile of mood states: technical update. North Tonawanda (NY): Multi-Health Systems; 2013. Crossref

25. Williams VS, Morlock RJ, Feltner D. Psychometric evaluation of a visual analog scale for the assessment of anxiety. Health Qual Life Outcomes 2010;8:57. Crossref

26. Facco E, Stellini E, Bacci C, et al. Validation of visual analogue scale for anxiety (VAS-A) in preanesthesia evaluation. Minerva Anestesiol 2013;79:1389-95.

27. Cumming GP, Klein S, Bolsover D, et al. The emotional burden of miscarriage for women and their partners: trajectories of anxiety and depression over 13 months. BJOG 2007;114:1138-45. Crossref

28. Virgara R, Maher C, Van Kessel G. The comorbidity of low back pelvic pain and risk of depression and anxiety in pregnancy in primiparous women. BMC Pregnancy Childbirth 2018;18:288. Crossref

29. Kong GW, Lok IH, Lam PM, Yip AS, Chung TK. Conflicting perceptions between health care professionals and patients on the psychological morbidity following miscarriage. Aust N Z J Obstet Gynaecol 2010;50:562-7. Crossref

30. Vitale SG, Laganà AS, Muscatello MR, et al. Psychopharmacotherapy in pregnancy and breastfeeding. Obstet Gynecol Surv 2016;71:721-33. Crossref

Anaemia is a global health issue that affects one-quarter of the world’s population; it is particularly prevalent among preschool-aged children in developing countries.1 Approximately 47.4% of preschool-aged children worldwide display anaemia.1 There are three categories of factors associated with anaemia: inherited disorders, infectious diseases, and micronutrient deficiencies.2 3 Among these factors, iron deficiency is the most common cause,4 especially in China.5 There is evidence that iron deficiency anaemia affects developmental potential in children.6 7

Anaemia prevalence among children in China, particularly in poor rural areas, is higher than that in developed countries.2 3 In the United States and the Netherlands, the rate is <10%.2 The rate in urban areas of China is <20%,8 9 whereas the prevalence in rural areas is more than double that in urban areas.10 11 12 Thus, there is a need for considerable effort from the Chinese Government to ensure that regional anaemia prevalence among children aged <5 years are below 10% by 2030.13

Few studies have examined factors associated with anaemia among children of different ages, particularly in poor rural areas of China. Previous studies have shown that anaemia may be associated with the demographic, social, and health characteristics of children and their families.14 15 16 17 Feeding practices have also been associated with anaemia in children.1 18 19 20 However, few studies have extensively analysed anaemia prevalence and associated factors among children of different ages in rural China.16 21 22 For example, one study explored risk factors for anaemia in children aged 0 to 5 months and those aged 6 to 36 months; however, the age ranges were excessively broad.14 In another study exploring risk factors for anaemia in children aged <36 months, stratified according to age, relatively few potential associated factors (eg, socio-demographic and illness characteristics) were considered; there was no consideration of other potential associated factors, such as complementary feeding.18

This study was therefore conducted to explore anaemia prevalence and risk factors among children aged 6 to 23 months in poor rural areas of China; analyses were performed focusing on the overall sample and with stratification according to age. Therefore, we established three objectives: to examine anaemia prevalence among children in the study area; to identify socio-demographic and illness characteristics associated with anaemia in children; and to explore feeding practices associated with anaemia in children.

This study was conducted in 22 nationally designated poverty-stricken counties (all of which are now out of poverty) within three prefectures in the Qinba Mountains area of northwest China. By the end of 2015 in the survey year, the total population of the sample area was 8 464 200, including a rural population of 4 716 100 (55.7%). The per capita income was 20 939 yuan, which was less than half of the national per capita income (42 359 yuan) in the same period in China.23 Sample villages and households were selected in two stages. First, from each of the 22 counties, all townships (ie, the middle level of administration between county and village) that met the criteria were selected to participate in the study, with two exceptions: the township in each county containing the county government (which represents the level of county development), as well as townships containing <800 people. In total, 115 of 400 townships were included in this study. Second, in each sample township, we selected random villages with ≥10 children. All children in our target age range (6-23 months) were enrolled in the study, including premature but not congenitally abnormal children; thus, we included 1694 children and their households. Because one prefecture did not survey feeding practices, the corresponding analysis only included 1210 participants from the other two sample prefectures. In total, 1132 participants (children and their households) fully completed the survey (response rate of 93.6%).

Survey data were collected in three waves in November 2015, April 2016, and February 2017. After identification of the primary caregiver responsible for a child’s diet and care, well-trained enumerators collected information through one-on-one questionnaire interviews with the primary caregiver.

First, specific components of socio-demographic and illness characteristics were recorded in the survey. The socio-demographic characteristics included the child’s age, sex, gestational age, and birth order; the primary caregiver’s identity; maternal education and age; and whether the family received social security support (ie, government welfare for the lowest income families nationwide). Illness characteristics comprised any history of fever, cold, or diarrhoea in the previous 2 weeks.

Second, detailed information regarding the child’s feeding practices was collected via dietary recall, using a series of questions based on the ‘Indicators for assessing infant feeding practices’ compiled by the World Health Organization (WHO).24 The following definitions were used: continued breastfeeding, proportion of children aged 6 to 23 months who had received breast milk during the previous day; any history of formula feeding, proportion of children who had ever been formula-fed; minimum dietary diversity, proportion of children aged 6 to 23 months who had consumed ≥4 of the 7 food groups under WHO’s classification24 during the previous day; minimum meal frequency, proportion of children aged 6 to 23 months who consumed a meal at a standard frequency during the previous day, considering their breastfeeding status (two times for breastfed infants aged 6 to 8 months, three times for breastfed children aged 9 to 23 months, and four times for non-breastfed children aged 6 to 23 months); minimum acceptable diet, proportion of children aged 6 to 23 months who consumed a meal that met standards for minimum dietary diversity and minimum meal frequency during the previous day; and consumption of iron-rich or iron-fortified foods, proportion of children who consumed iron-rich or iron-fortified foods specifically designed for children aged 6 to 23 months during the previous day.

Third, each child’s haemoglobin (Hb) concentration and anaemia status were assessed by trained nurses from the Xi’an Jiaotong University, who performed tests on fingertip blood samples collected from all children. These analyses were performed using the HemoCue Hb201 haemoglobin analyser (HemoCue Inc, Ängelholm, Sweden), which is accurate, rapid, and convenient for children in remote rural areas.11 15 21 25 Its measurement accuracy is 1 g/L.18 We confirmed that the sample villages’ altitudes were below 1000 m; therefore, no adjustments to measured Hb concentrations were required. Anaemia status was determined according to Hb concentration and divided into four categories: non-anaemic, Hb concentration ≥110 g/L; mild, 100-109 g/L; moderate, 70-99 g/L; and severe, <70 g/L.26 Children with severe anaemia were referred to a local hospital for treatment.

Statistical analysis was performed using STATA version 15.0 (Stata Corporation, College Station [TX], United States). The children’s socio-demographic and illness characteristics, feeding practices, and anaemia statuses were summarised using descriptive statistics. In bivariate analyses, P values for differences in mean Hb concentration between subgroups were estimated using t tests. The Pearson Chi squared test was also used to compare categorical variables between anaemia and non-anaemia groups. Multiple linear regression analyses were performed to identify covariates that were significantly associated with Hb concentration. Multiple logistic regression analysis was used to identify predictors of anaemia. The threshold for statistical significance was set at P<0.05.

Table 1 presents the socio-demographic and illness characteristics of the 1132 children. Of these, 51.0% were boys, 5.2% were born prematurely, and more than half were first-born (54.9%). Additionally, more than half of the primary caregivers (68.9%) were the children’s mothers; the remaining primary caregivers were the children’s grandmothers. Less than one-quarter of the children’s mothers (22.5%) had >9 years of education, and more than half of them (59.4%) were aged ≤28 years. Social security support was received by 11.9% of the participating families. Approximately half of the children (55.6%) had been sick (with fever, cold, or diarrhoea) in the previous 2 weeks.

Table 1 also presents the feeding practices of the children; notably, 29.8% and 86.6% of the children had continued breastfeeding and any history of formula feeding, respectively. With respect to complementary feeding, most children (80.9%) consumed iron-rich or iron-fortified foods; however, approximately 65.0% and 44.2% of the children met the standard requirements for minimum dietary diversity and meal frequency, respectively. Moreover, only 19.9% of the children met the standard requirement for a minimum acceptable diet. All children were divided into three age-groups: 6 to 11 months (n=343), 12 to 17 months (n=472), and 18 to 23 months (n=317).

Table 2 presents the children’s Hb concentrations and anaemia prevalence; the mean and standard deviation of their Hb concentration was 110.95 ± 0.42 g/L. Overall, 42.6% of the children had anaemia, including 21.6% with mild anaemia, 20.1% with moderate anaemia, and 0.8% with severe anaemia. A similar pattern was observed upon stratification according to age: few children had severe anaemia, and approximately one-quarter of children displayed mild or moderate anaemia in 6 to 11 months and 12 to 17 months age-groups.

As age increased across the groups (from 6-11 months to 12-17 months, and then to 18-23 months), the mean Hb concentration increased, whereas anaemia prevalence decreased. The mean and standard deviation Hb concentrations in the three groups (from youngest to oldest) were 106.85 ± 0.72 g/L, 111.10 ± 0.64 g/L, and 115.18 ± 0.78 g/L, respectively. Furthermore, children aged 6 to 11 months had the highest anaemia prevalence (53.6%), followed by children aged 12 to 17 months (43.4%) and then children aged 18 to 23 months (29.3%).

Table 3 shows the bivariate associations of Hb concentration/anaemia prevalence with the children’s socio-demographic and illness characteristics, stratified according to age. Among children aged 12 to 17 months, birth order and health status were significantly associated with Hb concentration/anaemia prevalence; however, the associations were not statistically significant in the other two age-groups or the overall sample. Among children aged 12 to 17 months, Hb concentrations were significantly higher in first-born children than in non-first-born children (P=0.020). Moreover, among children aged 12 to 17 months, children who had been sick in the previous 2 weeks were more likely to display anaemia, compared with children who had not been sick (P=0.029).

A similar trend was observed regarding the relationship of Hb concentration/anaemia prevalence with the primary caregiver; however, the only statistically significant result was observed in the overall sample. In summary, the Hb concentration was lower (P=0.003) and anaemia prevalence was higher (P=0.001) among children whose primary caregiver was their mother, compared with children who had a different primary caregiver. Furthermore, in the overall sample and all age-groups, there were no significant binary associations between the Hb concentration/anaemia prevalence and variables such as sex, premature birth, maternal education and age, or receipt of social security support.

Table 4 shows the bivariate associations of Hb concentration/anaemia prevalence with feeding practices. The associations varied among age-groups and in the overall sample. Children with any history of formula feeding had higher Hb concentrations and lower rates of anaemia, compared with children who had never received formula (both P<0.001); these differences were statistically significant in all age-groups. Children who had continued breastfeeding displayed lower Hb concentrations and higher rates of anaemia, compared with children who had stopped breastfeeding (both P<0.001); these differences were statistically significant among children aged 12 to 17 months (both Hb concentration and anaemia prevalence) and 18 to 23 months (anaemia prevalence only).

Additionally, observable complementary food–related variables were significantly associated with Hb concentration and anaemia prevalence. In the overall sample, children with feeding practices that met the minimum requirements for dietary diversity had significantly higher Hb concentrations (P<0.001) and lower rates of anaemia (P=0.005), compared with children whose feeding practices did not meet those requirements. Children with feeding practices that met the minimum meal frequency requirements had higher Hb concentrations (P=0.018), compared with children whose feeding practices did not meet those requirements. Regarding the consumption of iron-rich or iron-fortified foods, a significant positive association with Hb concentration and a significant negative association with anaemia prevalence was observed among children aged 12 to 17 months and in the overall sample (both P<0.001).

The results of multivariable analysis of the relationship between Hb concentration and anaemia prevalence are presented in Table 5. The initial multivariable model included variables related to socio-demographic and illness characteristics, continued breastfeeding, and any history of formula feeding; the results showed that Hb concentrations were significantly higher in first-born children (P=0.031) and significantly lower in children of younger mothers (P=0.032), but no factors were significantly associated with anaemia prevalence. Any history of formula feeding was positively associated with Hb concentration (P=0.031) and negatively associated with anaemia prevalence (odds ratio [OR]=0.59, 95% confidence interval [CI]=0.41-0.86; P=0.006), whereas continued breastfeeding was significantly negatively associated with Hb concentration (P=0.001) and positively associated with anaemia prevalence (OR=1.50, 95% CI=1.07-2.11; P=0.019). A subsequent multivariable model included socio-demographic and illness characteristics, as well as complementary food–related variables; the results showed that Hb concentration remained positively associated with first-born-child status (P=0.025) and younger maternal age (P=0.032), whereas consumption of iron-rich or iron-fortified foods was negatively associated with anaemia prevalence (OR=0.66, 95% CI=0.46-0.94; P=0.021). The final multivariable model included all variables; the results showed that continued breastfeeding was positively associated with anaemia prevalence (OR=1.75, 95% CI=1.21-2.51; P=0.003), whereas any history of formula feeding was negatively associated with anaemia prevalence (OR=0.57, 95% CI=0.38-0.87; P=0.010).

In this analysis of 1132 children aged 6 to 23 months in a poor rural area of China, we found that the anaemia prevalence was high in the overall sample, although it varied among age-groups. Bivariate analysis of socio-demographic characteristics, illness characteristics, and feeding practices revealed diverse risk factors among age-groups and in the overall sample. Additionally, multivariable analysis showed that feeding practice–related variables were risk factors for anaemia prevalence. Compared with complementary food–related variables, continued breastfeeding and any history of formula feeding had much greater impacts across age-groups.

Our findings revealed that 42.6% of children in the overall sample displayed anaemia, and anaemia prevalence among children in rural China varied according to age. According to WHO guidelines, anaemia prevalence exceeding 40% is a ‘severe public health problem’.26 Previous studies revealed anaemia prevalence among children in rural areas of central China (29.7%) and eastern China (24.2%)21 27; the prevalence was higher among children in our sample, indicating that urgent attention is needed regarding anaemia among children in rural areas of western China. Furthermore, our results showed that anaemia prevalence decreased with increasing age, consistent with previous reports.8 15 17 28 We found that anaemia prevalence was lower among children aged 18 to 23 months than among those aged 6 to 11 months or 12 to 17 months; this may have been related to the successful inclusion of complementary foods after 12 months of age. There is evidence that increasing iron intake from various foods contributes to a slow decrease in anaemia prevalence.26 Overall, our findings imply substantial differences in anaemia prevalence according to age; thus, analyses of anaemia in children, along with its risk factors, should consider the effect of age (in months).

Our bivariate analysis showed significant differences in risk factors for low Hb concentration and anaemia prevalence among children in the overall sample and in each age-group. These findings were consistent with the results of other studies regarding anaemia among children in China.21 22 In particular, a study of children aged 6 to 23 months showed that complementary feeding practices meeting the minimum dietary diversity requirement were negatively associated with anaemia prevalence among children aged 12 to 17 months; however, the association was not statistically significant among children aged 6 to 11 months or 18 to 23 months. Additionally, complementary feeding practices meeting the minimum meal frequency requirement were negatively associated with anaemia prevalence in all age-groups.22 Therefore, we conclude that the risk factors for anaemia prevalence in children differ according to age.

Our results also indicated that socio-demographic and illness characteristics were associated with anaemia prevalence among children in poor rural areas of China, consistent with previous findings.11 17 Specifically, birth order and a history of illness in the previous 2 weeks were statistically significant risk factors for anaemia in children aged 12 to 17 months. Regarding health status, previous studies revealed that anaemia is positively associated with a history of recurrent illness, such as diarrhoea or fever.11 19 We found that children who had been sick in the previous 2 weeks were more likely to display anaemia, presumably because they experienced a loss of appetite and had poor intestinal nutrient absorption.27 The child’s relationship with their primary caregiver was significantly associated with Hb concentration and anaemia prevalence in the overall sample. Previous studies showed greater dependence on breast milk among children whose primary caregiver was their mother; this dependence may lead to anaemia. Thus, the provision of adequate nutrition via complementary food is recommended.29

Bivariate and multivariable analyses showed that feeding practices (continued breastfeeding, any history of formula feeding, and consumption of iron-rich or iron-fortified foods) were associated with anaemia prevalence in poor rural areas of China. However, continued breastfeeding and any history of formula feeding had greater impacts on specific age-groups. Children who had continued breastfeeding displayed significantly lower Hb concentrations and higher rates of anaemia, both in the overall sample and among children aged 12 to 17 months or 18 to 23 months. These findings are consistent with the results of previous studies.30 31 32 Although the importance of breastfeeding for children before the age of 2 is widely recognised, empirical studies have shown that prolonged breastfeeding (ie, beyond 6 months of age) is positively associated with anaemia in children aged <2 years.31 32 Increases in total breastfeeding duration are associated with decreases in iron stores, implying late introduction or poor quality of complementary foods in children, as well as maternal anaemia.31 33 Accordingly, although there remains a need to encourage breastfeeding, careful monitoring of maternal and infant anaemia should be implemented, along with timely introduction of appropriate complementary foods to infants by 6 months of age; maternal diets and nutritional supplementation should also be improved.33 Children with any history of formula feeding had a higher Hb concentration and lower anaemia prevalence in each age-group, as well as the overall sample, consistent with previous findings.11 19 34 Formula feeding protects against anaemia in children, presumably because most commercially available formulas are fortified with micronutrients (eg, iron).27 Children with any history of formula feeding would have received additional iron, which have may helped to improve their anaemia status.11 Therefore, high-iron formulas are recommended for infants aged >6 months.35

In the overall sample, children with feeding practices that minimum dietary diversity standards and children who consumed iron-rich or iron-fortified foods were less likely to display anaemia. These results are consistent with the findings of studies in other rural areas of China.16 20 22 Regarding minimum dietary diversity, the WHO recommends that children aged 6 to 23 months receive a variety of foods to ensure that their nutrient requirements are met.36 A child’s needs with respect to the type and quantity of complementary foods increase with monthly age.37 Other studies have shown that the addition of complementary food in moderate amounts protects against anaemia.18 30 After 6 months of age, sources of iron for anaemia prevention are mainly derived from complementary foods.19 20 22 The consumption of iron-rich foods can reduce the risk of anaemia by improving iron storage and subsequent Hb production.19 The results of some studies have highlighted the importance of high-energy foods rich in iron, including beans, dark green leafy vegetables, meat, and viscera. These foods constitute sources of haem iron, which has better bioavailability.18 Therefore, caregivers should receive information concerning the importance of iron-rich complementary foods before they begin introducing complementary foods to their children.37

However, there is evidence that many children in rural China do not meet the standards for complementary feeding recommended by the WHO.18 22 24 Family income level substantially impacts nutritional intake.20 Although formula and complementary foods are widely available, they may not be prioritised in poor rural households.22 Because parents in such households often lack nutritional knowledge, they may assume that nutrient deficiency is unlikely; this belief can lead to inappropriate feeding in many children.20 Therefore, active intervention is needed; effective communication methods should be established to provide nutritional health knowledge and social support for family nutrition.

This study had several important limitations. First, we could not determine whether seasonal or temporal factors were associated with anaemia. Although we had some seasonal and temporal data regarding the three survey waves, key information was unavailable; thus, we could not confirm the findings of Luo et al.11 Second, although previous studies indicated that anaemia during pregnancy is a risk factor for anaemia in children,38 39 the present study lacked data regarding maternal anaemia during pregnancy; thus, we could not explore this relationship. Third, we only assessed any history of formula feeding, rather than ongoing formula feeding, which may have led to inaccurate results. Additional studies are needed to address these limitations.

Anaemia remains a severe public health problem among children aged 6 to 23 months in rural China. Continued breastfeeding was significantly positively associated with anaemia prevalence, whereas any history of formula feeding and the consumption of iron-rich or iron-fortified foods were significantly negatively associated with anaemia prevalence. Although we could not make causal inferences on the basis of findings in this cross-sectional study, our analysis provided key information concerning factors associated with anaemia prevalence among children of various ages in rural China; these findings will help to guide clinical practice and support policy formulation.

Concept or design: L Zeng, W Zheng, Q Gao.
Acquisition of data: K Du, A Yue.
Analysis or interpretation of data: L Zeng, Q Gao.
Drafting of the manuscript: W Zheng, A Yue, Q Gao.
Critical revision of the manuscript for important intellectual content: Q Gao, N Qiao.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

All authors have disclosed no conflicts of interest.

We thank the enumerators for their contribution to data collection.

This research was supported by the 111 Project (Grant No.: B16031), the National Social Science Foundation of China (Grant No.: 22BGL212), the National Natural Science Foundation of China (Grant No.: 72203134), and the Special Project of Philosophy and Social Science Research in Shaanxi Province (Grant No.: 2023QN0058). The funders had no role in study design, data collection/analysis/interpretation, or manuscript preparation.

This study protocol was approved by the Sichuan University Institutional Review Board of China (Protocol ID: 2013005-01). All caregivers of the children under investigation provided oral informed consent before participating in this study.

1. McLean E, Cogswell M, Egli I, Wojdyla D, de Benoist B. Worldwide prevalence of anaemia, WHO Vitamin and Mineral Nutrition Information System, 1993-2005. Public Health Nutr 2009;12:444-54. Crossref

2. Tolentino K, Friedman JF. An update on anemia in less developed countries. Am J Trop Med Hyg 2007;77:44-51.Crossref

3. Balarajan Y, Ramakrishnan U, Ozaltin E, Shankar AH, Subramanian SV. Anaemia in low-income and middle-income countries. Lancet 2011;378:2123-35. Crossref

4. World Health Organization. Anaemia. Available from: https://www.who.int/health-topics/anaemia#tab=tab_1. Accessed 17 May 2023.

5. Wang L, Sun Y, Liu B, et al. Is infant/toddler anemia a problem across rural China? A mixed-methods analysis. Int J Environ Res Public Health 2018;15:1825. Crossref

6. Lozoff B, Smith JB, Kaciroti N, Clark KM, Guevara S, Jimenez E. Functional significance of early-life iron deficiency: outcomes at 25 years. J Pediatr 2013;163:1260-6. Crossref

7. Deng X. Effects of iron deficiency anemia on physical development, intelligent behavior development and sleep quality of infants [in Chinese]. Pract Clin Med 2019;4:719-81.

8. Jin D, Zhuo Y, Hu L, Ruan D. Influence of iron deficiency anemia on intelligence of infants aged 6 to 36 months [in Chinese]. Chin J Woman Child Health Res 2019;30:281-3.

9. Zhang B. Analysis of the prevalence of anemia in infants and young children and the causes of transformation in a community in Shanghai [in Chinese]. Shanghai Med Pharm J 2019;40:61-2.

10. Wang L, Liang W, Zhang S, et al. Are infant/toddler developmental delays a problem across rural China? J Comp Econ 2019;47:458-69. Crossref

11. Luo R, Shi Y, Zhou H, et al. Anemia and feeding practices among infants in rural Shaanxi province in China. Nutrients 2014;6:5975-91. Crossref

12. Nie J, Yang J, Zhang L, Li Y, Yang J. Current situation and influencing factors of anemia in infants in poor rural areas in western China. J East China Norm Univ Sci 2019;37:58-69.

13. The State Council Information Office of the People’s Republic of China. Circular of the General Office of the State Council on the Issuance of China’s Food and Nutrition Development Program (2017-2030) [in Chinese]. State Office 2017. Available from: http://www.scio.gov.cn/xwfbh/xwbfbh/wqfbh/42311/44583/xgzc44589/Document/1695253/1695253.htm Accessed 9 Sep 2021.

14. Gao W, Yan H, Wang D, Dang S, Pei L. Severity of anemia among children under 36 months old in rural western China. PLoS One 2013;8:e62883. Crossref

15. Chen L, Li Y, Li Y, Yang F, Bi X. Iron deficiency anemia status and influencing factors among children aged 6 to 23 months in rural minority areas in Yunnan [in Chinese]. Chin J Child Health Care 2017;25:85-7.

16. Huang Y, Wang L, Huo J, et al. Prevalence and causes of anaemia in children aged 6-23 months in rural Qinghai, China: findings from a cross-sectional study. BMJ Open 2019;9:e031021. Crossref

17. Yang Y, Liu L, Yang W, Zhang S, Wang Y. State and influencing factors for anemia of infants aged 0-18 months in rural area of Shaanxi province [in Chinese]. Matern Child Health Care China 2013;28:1897-9.

18. Luo RF, Liang X, Liu CF, Zhang LX, Yue A. Risk factors for anemia in infants aged 6-12 months from rural areas of southern Shaanxi Province, China [in Chinese]. Chin J Contemp Pediatr 2016;18:736-41.

19. Sun S, Wu Y, Chen Y, Hao G, Zhou H. Analysis of the influencing factors of infant anemia in underdeveloped rural areas of southern Shaanxi [in Chinese]. Chin J Dis Control Prev 2015;19:1138-41.

20. Zhou X, Fang JQ, Luo JY, et al. Prevalence and influencing factors of anemia among 6-23 months old children in poor rural areas [in Chinese]. Chin J Public Health 2017;33:1302-5.

21. Xin QQ, Chen BW, Yin DL, et al. Prevalence of anemia and its risk factors among children under 36 months old in China. J Trop Pediatr 2017;63:36-42. Crossref

22. Hipgrave DB, Fu X, Zhou H, et al. Poor complementary feeding practices and high anaemia prevalence among infants and young children in rural central and western China. Eur J Clin Nutr 2014;68:916-24. Crossref

23. Shaanxi Provincial Bureau of Statistics. Shaanxi statistical yearbook [in Chinese]. 2016. Available from: http://tjj.shaanxi.gov.cn/tjsj/ndsj/tjnj/sxtjnj/index.html?2016. Accessed 17 May 2023.

24. World Health Organization. Indicators for assessing infant and young child feeding practices (Part I—definitions). Conclusions of a consensus meeting held 6-8 November 2007 in Washington, DC, USA. 2008. Available from: https://apps.who.int/iris/bitstream/handle/10665/43895/9789241596664_eng.pdf;jsessionid=96860366C0058393F12BC11CE25498BA?sequence=1. Accessed 23 Dec 2021.

25. Seguin P, Kleiber A, Chanavaz C, Morcet J, Mallédant Y. Determination of capillary hemoglobin levels using the HemoCue system in intensive care patients. J Crit Care 2011;26:423-7. Crossref

26. World Health Organization. Haemoglobin concentrations for the diagnosis of anaemia and assessment of severity. 2011. Available from: https://apps.who.int/iris/bitstream/handle/10665/85839/WHO_NMH_NHD_MNM_11.1_eng.pdf?sequence=22&isAllowed=y. Accessed 4 Jul 2023.

27. Huang Z, Jiang FX, Li J, Jiang D, Xiao TG, Zeng JH. Prevalence and risk factors of anemia among children aged 6-23 months in Huaihua, Hunan Province. BMC Public Health 2018;18:1267. Crossref

28. Huo J, Sun J, Fang Z, et al. Effect of home-based complementary food fortification on prevalence of anemia among infants and young children aged 6 to 23 months in poor rural regions of China. Food Nutr Bull 2015;36:405-14. Crossref

29. Li H, Xiao J, Liao M, et al. Anemia prevalence, severity and associated factors among children aged 6-71 months in rural Hunan Province, China: a community-based cross-sectional study. BMC Public Health 2020;20:989. Crossref

30. Hopkins D, Emmett P, Steer C, Rogers I, Noble S, Emond A. Infant feeding in the second 6 months of life related to iron status: an observational study. Arch Dis Child 2007;92:850-4. Crossref

31. Hu S, Tan H, Peng A, et al. Disparity of anemia prevalence and associated factors among rural to urban migrant and the local children under two years old: a population based cross-sectional study in Pinghu, China. BMC Public Health 2014;14:601. Crossref

32. Wu LX, Wang PW, Liu S, Li SY, Yan H, Zuo D. Analysis on breast feeding and anemia status among infants under 2 years old in rural areas of Hubei province [in Chinese]. Chin J Child Health Care 2013;21:1175-7.

33. Meinzen-Derr JK, Guerrero ML, Altaye M, Ortega-Gallegos H, Ruiz-Palacios GM, Morrow AL. Risk of infant anemia is associated with exclusive breast-feeding and maternal anemia in a Mexican cohort. J Nutr 2006;136:452-8. Crossref

34. Dong C, Ge P, Zhang C, et al. Effects of different feeding practices at 0 to 6 months and living economic conditions on anemia prevalence of infants and young children [in Chinese]. J Hyg Res 2013;42:596-9.

35. Vendt N, Grünberg H, Leedo S, Tillmann V, Talvik T. Prevalence and causes of iron deficiency anemias in infants aged 9 to 12 months in Estonia. Medicina (Kaunas) 2007;43:947-52. Crossref

36. World Health Organization. Indicators for assessing infant and young child feeding practices: definitions and measurement methods. 2021. Available from: https://www.who.int/publications/i/item/9789240018389. Accessed 20 Sep 2021.

37. Lu P, Wang J, Jiang W, et al. Feeding status of 0~23-month-old infants in poor rural areas of Gansu Province from 2018 to 2019 [in Chinese]. Wei Sheng Yan Jiu 2020;49:731-43.

38. Wang L. Investigation on risk factors of iron deficiency anemia in infants [in Chinese]. Chin J Woman Child Health Res 2017;(S4):434-5.

39. Shukla AK, Srivastava S, Verma G. Effect of maternal anemia on the status of iron stores in infants: a cohort study. J Family Community Med 2019;26:118-22. Crossref